xref: /petsc/src/mat/impls/sbaij/seq/sbaijfact.c (revision cf12db733d96eb19aec68f4bc970b9b4e51956bf)
1 #define PETSCMAT_DLL
2 
3 #include "src/mat/impls/baij/seq/baij.h"
4 #include "src/mat/impls/sbaij/seq/sbaij.h"
5 #include "src/inline/ilu.h"
6 #include "include/petscis.h"
7 
8 #if !defined(PETSC_USE_COMPLEX)
9 /*
10   input:
11    F -- numeric factor
12   output:
13    nneg, nzero, npos: matrix inertia
14 */
15 
16 #undef __FUNCT__
17 #define __FUNCT__ "MatGetInertia_SeqSBAIJ"
18 PetscErrorCode MatGetInertia_SeqSBAIJ(Mat F,PetscInt *nneig,PetscInt *nzero,PetscInt *npos)
19 {
20   Mat_SeqSBAIJ *fact_ptr = (Mat_SeqSBAIJ*)F->data;
21   MatScalar    *dd = fact_ptr->a;
22   PetscInt     mbs=fact_ptr->mbs,bs=F->rmap.bs,i,nneig_tmp,npos_tmp,*fi = fact_ptr->i;
23 
24   PetscFunctionBegin;
25   if (bs != 1) SETERRQ1(PETSC_ERR_SUP,"No support for bs: %D >1 yet",bs);
26   nneig_tmp = 0; npos_tmp = 0;
27   for (i=0; i<mbs; i++){
28     if (PetscRealPart(dd[*fi]) > 0.0){
29       npos_tmp++;
30     } else if (PetscRealPart(dd[*fi]) < 0.0){
31       nneig_tmp++;
32     }
33     fi++;
34   }
35   if (nneig) *nneig = nneig_tmp;
36   if (npos)  *npos  = npos_tmp;
37   if (nzero) *nzero = mbs - nneig_tmp - npos_tmp;
38 
39   PetscFunctionReturn(0);
40 }
41 #endif /* !defined(PETSC_USE_COMPLEX) */
42 
43 /*
44   Symbolic U^T*D*U factorization for SBAIJ format. Modified from SSF of YSMP.
45   Use Modified Sparse Row (MSR) storage for u and ju. See page 85, "Iterative Methods ..." by Saad.
46 */
47 #undef __FUNCT__
48 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqSBAIJ_MSR"
49 PetscErrorCode MatCholeskyFactorSymbolic_SeqSBAIJ_MSR(Mat A,IS perm,MatFactorInfo *info,Mat *B)
50 {
51   Mat_SeqSBAIJ   *a = (Mat_SeqSBAIJ*)A->data,*b;
52   PetscErrorCode ierr;
53   PetscInt       *rip,i,mbs = a->mbs,*ai,*aj;
54   PetscInt       *jutmp,bs = A->rmap.bs,bs2=a->bs2;
55   PetscInt       m,reallocs = 0,prow;
56   PetscInt       *jl,*q,jmin,jmax,juidx,nzk,qm,*iu,*ju,k,j,vj,umax,maxadd;
57   PetscReal      f = info->fill;
58   PetscTruth     perm_identity;
59 
60   PetscFunctionBegin;
61   /* check whether perm is the identity mapping */
62   ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr);
63   ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr);
64 
65   if (perm_identity){ /* without permutation */
66     a->permute = PETSC_FALSE;
67     ai = a->i; aj = a->j;
68   } else {            /* non-trivial permutation */
69     a->permute = PETSC_TRUE;
70     ierr = MatReorderingSeqSBAIJ(A,perm);CHKERRQ(ierr);
71     ai = a->inew; aj = a->jnew;
72   }
73 
74   /* initialization */
75   ierr  = PetscMalloc((mbs+1)*sizeof(PetscInt),&iu);CHKERRQ(ierr);
76   umax  = (PetscInt)(f*ai[mbs] + 1); umax += mbs + 1;
77   ierr  = PetscMalloc(umax*sizeof(PetscInt),&ju);CHKERRQ(ierr);
78   iu[0] = mbs+1;
79   juidx = mbs + 1; /* index for ju */
80   ierr  = PetscMalloc(2*mbs*sizeof(PetscInt),&jl);CHKERRQ(ierr); /* linked list for pivot row */
81   q     = jl + mbs;   /* linked list for col index */
82   for (i=0; i<mbs; i++){
83     jl[i] = mbs;
84     q[i] = 0;
85   }
86 
87   /* for each row k */
88   for (k=0; k<mbs; k++){
89     for (i=0; i<mbs; i++) q[i] = 0;  /* to be removed! */
90     nzk  = 0; /* num. of nz blocks in k-th block row with diagonal block excluded */
91     q[k] = mbs;
92     /* initialize nonzero structure of k-th row to row rip[k] of A */
93     jmin = ai[rip[k]] +1; /* exclude diag[k] */
94     jmax = ai[rip[k]+1];
95     for (j=jmin; j<jmax; j++){
96       vj = rip[aj[j]]; /* col. value */
97       if(vj > k){
98         qm = k;
99         do {
100           m  = qm; qm = q[m];
101         } while(qm < vj);
102         if (qm == vj) {
103           SETERRQ(PETSC_ERR_PLIB,"Duplicate entry in A\n");
104         }
105         nzk++;
106         q[m]  = vj;
107         q[vj] = qm;
108       } /* if(vj > k) */
109     } /* for (j=jmin; j<jmax; j++) */
110 
111     /* modify nonzero structure of k-th row by computing fill-in
112        for each row i to be merged in */
113     prow = k;
114     prow = jl[prow]; /* next pivot row (== mbs for symbolic factorization) */
115 
116     while (prow < k){
117       /* merge row prow into k-th row */
118       jmin = iu[prow] + 1; jmax = iu[prow+1];
119       qm = k;
120       for (j=jmin; j<jmax; j++){
121         vj = ju[j];
122         do {
123           m = qm; qm = q[m];
124         } while (qm < vj);
125         if (qm != vj){
126          nzk++; q[m] = vj; q[vj] = qm; qm = vj;
127         }
128       }
129       prow = jl[prow]; /* next pivot row */
130     }
131 
132     /* add k to row list for first nonzero element in k-th row */
133     if (nzk > 0){
134       i = q[k]; /* col value of first nonzero element in U(k, k+1:mbs-1) */
135       jl[k] = jl[i]; jl[i] = k;
136     }
137     iu[k+1] = iu[k] + nzk;
138 
139     /* allocate more space to ju if needed */
140     if (iu[k+1] > umax) {
141       /* estimate how much additional space we will need */
142       /* use the strategy suggested by David Hysom <hysom@perch-t.icase.edu> */
143       /* just double the memory each time */
144       maxadd = umax;
145       if (maxadd < nzk) maxadd = (mbs-k)*(nzk+1)/2;
146       umax += maxadd;
147 
148       /* allocate a longer ju */
149       ierr = PetscMalloc(umax*sizeof(PetscInt),&jutmp);CHKERRQ(ierr);
150       ierr = PetscMemcpy(jutmp,ju,iu[k]*sizeof(PetscInt));CHKERRQ(ierr);
151       ierr = PetscFree(ju);CHKERRQ(ierr);
152       ju   = jutmp;
153       reallocs++; /* count how many times we realloc */
154     }
155 
156     /* save nonzero structure of k-th row in ju */
157     i=k;
158     while (nzk --) {
159       i           = q[i];
160       ju[juidx++] = i;
161     }
162   }
163 
164 #if defined(PETSC_USE_INFO)
165   if (ai[mbs] != 0) {
166     PetscReal af = ((PetscReal)iu[mbs])/((PetscReal)ai[mbs]);
167     ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,f,af);CHKERRQ(ierr);
168     ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr);
169     ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G);\n",af);CHKERRQ(ierr);
170     ierr = PetscInfo(A,"for best performance.\n");CHKERRQ(ierr);
171   } else {
172     ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr);
173   }
174 #endif
175 
176   ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr);
177   ierr = PetscFree(jl);CHKERRQ(ierr);
178 
179   /* put together the new matrix */
180   ierr = MatCreate(((PetscObject)A)->comm,B);CHKERRQ(ierr);
181   ierr = MatSetSizes(*B,bs*mbs,bs*mbs,bs*mbs,bs*mbs);CHKERRQ(ierr);
182   ierr = MatSetType(*B,((PetscObject)A)->type_name);CHKERRQ(ierr);
183   ierr = MatSeqSBAIJSetPreallocation_SeqSBAIJ(*B,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
184 
185   /* ierr = PetscLogObjectParent(*B,iperm);CHKERRQ(ierr); */
186   b = (Mat_SeqSBAIJ*)(*B)->data;
187   b->singlemalloc = PETSC_FALSE;
188   b->free_a       = PETSC_TRUE;
189   b->free_ij       = PETSC_TRUE;
190   ierr = PetscMalloc((iu[mbs]+1)*sizeof(MatScalar)*bs2,&b->a);CHKERRQ(ierr);
191   b->j    = ju;
192   b->i    = iu;
193   b->diag = 0;
194   b->ilen = 0;
195   b->imax = 0;
196   b->row  = perm;
197   b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */
198   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
199   b->icol = perm;
200   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
201   ierr    = PetscMalloc((bs*mbs+bs)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr);
202   /* In b structure:  Free imax, ilen, old a, old j.
203      Allocate idnew, solve_work, new a, new j */
204   ierr = PetscLogObjectMemory(*B,(iu[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr);
205   b->maxnz = b->nz = iu[mbs];
206 
207   (*B)->factor                 = MAT_FACTOR_CHOLESKY;
208   (*B)->info.factor_mallocs    = reallocs;
209   (*B)->info.fill_ratio_given  = f;
210   if (ai[mbs] != 0) {
211     (*B)->info.fill_ratio_needed = ((PetscReal)iu[mbs])/((PetscReal)ai[mbs]);
212   } else {
213     (*B)->info.fill_ratio_needed = 0.0;
214   }
215 
216   if (perm_identity){
217     switch (bs) {
218       case 1:
219         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering;
220         (*B)->ops->solve                 = MatSolve_SeqSBAIJ_1_NaturalOrdering;
221         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_1_NaturalOrdering;
222         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_1_NaturalOrdering;
223         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=1\n");CHKERRQ(ierr);
224         break;
225       case 2:
226         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
227         (*B)->ops->solve           = MatSolve_SeqSBAIJ_2_NaturalOrdering;
228         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_2_NaturalOrdering;
229         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_2_NaturalOrdering;
230         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=2\n");CHKERRQ(ierr);
231         break;
232       case 3:
233         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
234         (*B)->ops->solve           = MatSolve_SeqSBAIJ_3_NaturalOrdering;
235         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_3_NaturalOrdering;
236         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_3_NaturalOrdering;
237         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=3\n");CHKERRQ(ierr);
238         break;
239       case 4:
240         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
241         (*B)->ops->solve           = MatSolve_SeqSBAIJ_4_NaturalOrdering;
242         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_4_NaturalOrdering;
243         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_4_NaturalOrdering;
244         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=4\n");CHKERRQ(ierr);
245         break;
246       case 5:
247         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
248         (*B)->ops->solve           = MatSolve_SeqSBAIJ_5_NaturalOrdering;
249         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_5_NaturalOrdering;
250         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_5_NaturalOrdering;
251         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=5\n");CHKERRQ(ierr);
252         break;
253       case 6:
254         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
255         (*B)->ops->solve           = MatSolve_SeqSBAIJ_6_NaturalOrdering;
256         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_6_NaturalOrdering;
257         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_6_NaturalOrdering;
258         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=6\n");CHKERRQ(ierr);
259         break;
260       case 7:
261         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
262         (*B)->ops->solve           = MatSolve_SeqSBAIJ_7_NaturalOrdering;
263         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_7_NaturalOrdering;
264         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_7_NaturalOrdering;
265         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=7\n");CHKERRQ(ierr);
266       break;
267       default:
268         (*B)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
269         (*B)->ops->solve           = MatSolve_SeqSBAIJ_N_NaturalOrdering;
270         (*B)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_N_NaturalOrdering;
271         (*B)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_N_NaturalOrdering;
272         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS>7\n");CHKERRQ(ierr);
273       break;
274     }
275   }
276   PetscFunctionReturn(0);
277 }
278 /*
279     Symbolic U^T*D*U factorization for SBAIJ format.
280 */
281 #include "petscbt.h"
282 #include "src/mat/utils/freespace.h"
283 #undef __FUNCT__
284 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqSBAIJ"
285 PetscErrorCode MatCholeskyFactorSymbolic_SeqSBAIJ(Mat A,IS perm,MatFactorInfo *info,Mat *fact)
286 {
287   Mat_SeqSBAIJ       *a = (Mat_SeqSBAIJ*)A->data;
288   Mat_SeqSBAIJ       *b;
289   PetscErrorCode     ierr;
290   PetscTruth         perm_identity,missing;
291   PetscReal          fill = info->fill;
292   PetscInt           *rip,i,mbs=a->mbs,bs=A->rmap.bs,*ai,*aj,reallocs=0,prow,d;
293   PetscInt           *jl,jmin,jmax,nzk,*ui,k,j,*il,nextprow;
294   PetscInt           nlnk,*lnk,ncols,*cols,*uj,**ui_ptr,*uj_ptr;
295   PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL;
296   PetscBT            lnkbt;
297 
298   PetscFunctionBegin;
299   ierr = MatMissingDiagonal(A,&missing,&d);CHKERRQ(ierr);
300   if (missing) SETERRQ1(PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry %D",d);
301 
302   /*
303    This code originally uses Modified Sparse Row (MSR) storage
304    (see page 85, "Iterative Methods ..." by Saad) for the output matrix B - bad choise!
305    Then it is rewritten so the factor B takes seqsbaij format. However the associated
306    MatCholeskyFactorNumeric_() have not been modified for the cases of bs>1 or !perm_identity,
307    thus the original code in MSR format is still used for these cases.
308    The code below should replace MatCholeskyFactorSymbolic_SeqSBAIJ_MSR() whenever
309    MatCholeskyFactorNumeric_() is modified for using sbaij symbolic factor.
310   */
311   if (bs > 1){
312     ierr = MatCholeskyFactorSymbolic_SeqSBAIJ_MSR(A,perm,info,fact);CHKERRQ(ierr);
313     PetscFunctionReturn(0);
314   }
315 
316   /* check whether perm is the identity mapping */
317   ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr);
318 
319   if (perm_identity){
320     a->permute = PETSC_FALSE;
321     ai = a->i; aj = a->j;
322   } else {
323     SETERRQ(PETSC_ERR_SUP,"Matrix reordering is not supported for sbaij matrix. Use aij format");
324     /* There are bugs for reordeing. Needs further work.
325        MatReordering for sbaij cannot be efficient. User should use aij formt! */
326     a->permute = PETSC_TRUE;
327     ierr = MatReorderingSeqSBAIJ(A,perm);CHKERRQ(ierr);
328     ai = a->inew; aj = a->jnew;
329   }
330   ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr);
331 
332   /* initialization */
333   ierr  = PetscMalloc((mbs+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr);
334   ui[0] = 0;
335 
336   /* jl: linked list for storing indices of the pivot rows
337      il: il[i] points to the 1st nonzero entry of U(i,k:mbs-1) */
338   ierr = PetscMalloc((3*mbs+1)*sizeof(PetscInt)+mbs*sizeof(PetscInt*),&jl);CHKERRQ(ierr);
339   il     = jl + mbs;
340   cols   = il + mbs;
341   ui_ptr = (PetscInt**)(cols + mbs);
342 
343   for (i=0; i<mbs; i++){
344     jl[i] = mbs; il[i] = 0;
345   }
346 
347   /* create and initialize a linked list for storing column indices of the active row k */
348   nlnk = mbs + 1;
349   ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
350 
351   /* initial FreeSpace size is fill*(ai[mbs]+1) */
352   ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[mbs]+1)),&free_space);CHKERRQ(ierr);
353   current_space = free_space;
354 
355   for (k=0; k<mbs; k++){  /* for each active row k */
356     /* initialize lnk by the column indices of row rip[k] of A */
357     nzk   = 0;
358     ncols = ai[rip[k]+1] - ai[rip[k]];
359     for (j=0; j<ncols; j++){
360       i = *(aj + ai[rip[k]] + j);
361       cols[j] = rip[i];
362     }
363     ierr = PetscLLAdd(ncols,cols,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
364     nzk += nlnk;
365 
366     /* update lnk by computing fill-in for each pivot row to be merged in */
367     prow = jl[k]; /* 1st pivot row */
368 
369     while (prow < k){
370       nextprow = jl[prow];
371       /* merge prow into k-th row */
372       jmin = il[prow] + 1;  /* index of the 2nd nzero entry in U(prow,k:mbs-1) */
373       jmax = ui[prow+1];
374       ncols = jmax-jmin;
375       uj_ptr = ui_ptr[prow] + jmin - ui[prow]; /* points to the 2nd nzero entry in U(prow,k:mbs-1) */
376       ierr = PetscLLAddSorted(ncols,uj_ptr,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
377       nzk += nlnk;
378 
379       /* update il and jl for prow */
380       if (jmin < jmax){
381         il[prow] = jmin;
382         j = *uj_ptr; jl[prow] = jl[j]; jl[j] = prow;
383       }
384       prow = nextprow;
385     }
386 
387     /* if free space is not available, make more free space */
388     if (current_space->local_remaining<nzk) {
389       i = mbs - k + 1; /* num of unfactored rows */
390       i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */
391       ierr = PetscFreeSpaceGet(i,&current_space);CHKERRQ(ierr);
392       reallocs++;
393     }
394 
395     /* copy data into free space, then initialize lnk */
396     ierr = PetscLLClean(mbs,mbs,nzk,lnk,current_space->array,lnkbt);CHKERRQ(ierr);
397 
398     /* add the k-th row into il and jl */
399     if (nzk-1 > 0){
400       i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:mbs-1) */
401       jl[k] = jl[i]; jl[i] = k;
402       il[k] = ui[k] + 1;
403     }
404     ui_ptr[k] = current_space->array;
405     current_space->array           += nzk;
406     current_space->local_used      += nzk;
407     current_space->local_remaining -= nzk;
408 
409     ui[k+1] = ui[k] + nzk;
410   }
411 
412 #if defined(PETSC_USE_INFO)
413   if (ai[mbs] != 0) {
414     PetscReal af = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]);
415     ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,fill,af);CHKERRQ(ierr);
416     ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr);
417     ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G) for best performance.\n",af);CHKERRQ(ierr);
418   } else {
419     ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr);
420   }
421 #endif
422 
423   ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr);
424   ierr = PetscFree(jl);CHKERRQ(ierr);
425 
426   /* destroy list of free space and other temporary array(s) */
427   ierr = PetscMalloc((ui[mbs]+1)*sizeof(PetscInt),&uj);CHKERRQ(ierr);
428   ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr);
429   ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr);
430 
431   /* put together the new matrix in MATSEQSBAIJ format */
432   ierr = MatSeqSBAIJSetPreallocation_SeqSBAIJ(*fact,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
433 
434   b = (Mat_SeqSBAIJ*)(*fact)->data;
435   b->singlemalloc = PETSC_FALSE;
436   b->free_a       = PETSC_TRUE;
437   b->free_ij      = PETSC_TRUE;
438   ierr = PetscMalloc((ui[mbs]+1)*sizeof(MatScalar),&b->a);CHKERRQ(ierr);
439   b->j    = uj;
440   b->i    = ui;
441   b->diag = 0;
442   b->ilen = 0;
443   b->imax = 0;
444   b->row  = perm;
445   b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */
446   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
447   b->icol = perm;
448   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
449   ierr    = PetscMalloc((mbs+1)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr);
450   ierr    = PetscLogObjectMemory(*fact,(ui[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr);
451   b->maxnz = b->nz = ui[mbs];
452 
453   (*fact)->factor                 = MAT_FACTOR_CHOLESKY;
454   (*fact)->info.factor_mallocs    = reallocs;
455   (*fact)->info.fill_ratio_given  = fill;
456   if (ai[mbs] != 0) {
457     (*fact)->info.fill_ratio_needed = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]);
458   } else {
459     (*fact)->info.fill_ratio_needed = 0.0;
460   }
461 
462   if (perm_identity){
463     switch (bs) {
464       case 1:
465         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering;
466         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_1_NaturalOrdering;
467         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_1_NaturalOrdering;
468         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_1_NaturalOrdering;
469         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=1\n");CHKERRQ(ierr);
470         break;
471       case 2:
472         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
473         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_2_NaturalOrdering;
474         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_2_NaturalOrdering;
475         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_2_NaturalOrdering;
476         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=2\n");CHKERRQ(ierr);
477         break;
478       case 3:
479         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
480         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_3_NaturalOrdering;
481         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_3_NaturalOrdering;
482         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_3_NaturalOrdering;
483         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=3\n");CHKERRQ(ierr);
484         break;
485       case 4:
486         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
487         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_4_NaturalOrdering;
488         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_4_NaturalOrdering;
489         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_4_NaturalOrdering;
490         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=4\n");CHKERRQ(ierr);
491         break;
492       case 5:
493         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
494         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_5_NaturalOrdering;
495         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_5_NaturalOrdering;
496         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_5_NaturalOrdering;
497         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=5\n");CHKERRQ(ierr);
498         break;
499       case 6:
500         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
501         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_6_NaturalOrdering;
502         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_6_NaturalOrdering;
503         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_6_NaturalOrdering;
504         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=6\n");CHKERRQ(ierr);
505         break;
506       case 7:
507         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
508         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_7_NaturalOrdering;
509         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_7_NaturalOrdering;
510         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_7_NaturalOrdering;
511         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS=7\n");CHKERRQ(ierr);
512       break;
513       default:
514         (*fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
515         (*fact)->ops->solve           = MatSolve_SeqSBAIJ_N_NaturalOrdering;
516         (*fact)->ops->forwardsolve    = MatForwardSolve_SeqSBAIJ_N_NaturalOrdering;
517         (*fact)->ops->backwardsolve   = MatBackwardSolve_SeqSBAIJ_N_NaturalOrdering;
518         ierr = PetscInfo(A,"Using special in-place natural ordering factor and solve BS>7\n");CHKERRQ(ierr);
519       break;
520     }
521   }
522   PetscFunctionReturn(0);
523 }
524 #undef __FUNCT__
525 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_N"
526 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_N(Mat A,MatFactorInfo *info,Mat *B)
527 {
528   Mat            C = *B;
529   Mat_SeqSBAIJ   *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ *)C->data;
530   IS             perm = b->row;
531   PetscErrorCode ierr;
532   PetscInt       *perm_ptr,i,j,mbs=a->mbs,*bi=b->i,*bj=b->j;
533   PetscInt       *ai,*aj,*a2anew,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili;
534   PetscInt       bs=A->rmap.bs,bs2 = a->bs2,bslog = 0;
535   MatScalar      *ba = b->a,*aa,*ap,*dk,*uik;
536   MatScalar      *u,*diag,*rtmp,*rtmp_ptr;
537   MatScalar      *work;
538   PetscInt       *pivots;
539 
540   PetscFunctionBegin;
541   /* initialization */
542   ierr = PetscMalloc(bs2*mbs*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
543   ierr = PetscMemzero(rtmp,bs2*mbs*sizeof(MatScalar));CHKERRQ(ierr);
544   ierr = PetscMalloc(2*mbs*sizeof(PetscInt),&il);CHKERRQ(ierr);
545   jl   = il + mbs;
546   for (i=0; i<mbs; i++) {
547     jl[i] = mbs; il[0] = 0;
548   }
549   ierr = PetscMalloc((2*bs2+bs)*sizeof(MatScalar),&dk);CHKERRQ(ierr);
550   uik  = dk + bs2;
551   work = uik + bs2;
552   ierr = PetscMalloc(bs*sizeof(PetscInt),&pivots);CHKERRQ(ierr);
553 
554   ierr  = ISGetIndices(perm,&perm_ptr);CHKERRQ(ierr);
555 
556   /* check permutation */
557   if (!a->permute){
558     ai = a->i; aj = a->j; aa = a->a;
559   } else {
560     ai   = a->inew; aj = a->jnew;
561     ierr = PetscMalloc(bs2*ai[mbs]*sizeof(MatScalar),&aa);CHKERRQ(ierr);
562     ierr = PetscMemcpy(aa,a->a,bs2*ai[mbs]*sizeof(MatScalar));CHKERRQ(ierr);
563     ierr = PetscMalloc(ai[mbs]*sizeof(PetscInt),&a2anew);CHKERRQ(ierr);
564     ierr = PetscMemcpy(a2anew,a->a2anew,(ai[mbs])*sizeof(PetscInt));CHKERRQ(ierr);
565 
566     /* flops in while loop */
567     bslog = 2*bs*bs2;
568 
569     for (i=0; i<mbs; i++){
570       jmin = ai[i]; jmax = ai[i+1];
571       for (j=jmin; j<jmax; j++){
572         while (a2anew[j] != j){
573           k = a2anew[j]; a2anew[j] = a2anew[k]; a2anew[k] = k;
574           for (k1=0; k1<bs2; k1++){
575             dk[k1]       = aa[k*bs2+k1];
576             aa[k*bs2+k1] = aa[j*bs2+k1];
577             aa[j*bs2+k1] = dk[k1];
578           }
579         }
580         /* transform columnoriented blocks that lie in the lower triangle to roworiented blocks */
581         if (i > aj[j]){
582           /* printf("change orientation, row: %d, col: %d\n",i,aj[j]); */
583           ap = aa + j*bs2;                     /* ptr to the beginning of j-th block of aa */
584           for (k=0; k<bs2; k++) dk[k] = ap[k]; /* dk <- j-th block of aa */
585           for (k=0; k<bs; k++){               /* j-th block of aa <- dk^T */
586             for (k1=0; k1<bs; k1++) *ap++ = dk[k + bs*k1];
587           }
588         }
589       }
590     }
591     ierr = PetscFree(a2anew);CHKERRQ(ierr);
592   }
593 
594   /* for each row k */
595   for (k = 0; k<mbs; k++){
596 
597     /*initialize k-th row with elements nonzero in row perm(k) of A */
598     jmin = ai[perm_ptr[k]]; jmax = ai[perm_ptr[k]+1];
599 
600     ap = aa + jmin*bs2;
601     for (j = jmin; j < jmax; j++){
602       vj = perm_ptr[aj[j]];         /* block col. index */
603       rtmp_ptr = rtmp + vj*bs2;
604       for (i=0; i<bs2; i++) *rtmp_ptr++ = *ap++;
605     }
606 
607     /* modify k-th row by adding in those rows i with U(i,k) != 0 */
608     ierr = PetscMemcpy(dk,rtmp+k*bs2,bs2*sizeof(MatScalar));CHKERRQ(ierr);
609     i = jl[k]; /* first row to be added to k_th row  */
610 
611     while (i < k){
612       nexti = jl[i]; /* next row to be added to k_th row */
613 
614       /* compute multiplier */
615       ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
616 
617       /* uik = -inv(Di)*U_bar(i,k) */
618       diag = ba + i*bs2;
619       u    = ba + ili*bs2;
620       ierr = PetscMemzero(uik,bs2*sizeof(MatScalar));CHKERRQ(ierr);
621       Kernel_A_gets_A_minus_B_times_C(bs,uik,diag,u);
622 
623       /* update D(k) += -U(i,k)^T * U_bar(i,k) */
624       Kernel_A_gets_A_plus_Btranspose_times_C(bs,dk,uik,u);
625       ierr = PetscLogFlops(bslog*2);CHKERRQ(ierr);
626 
627       /* update -U(i,k) */
628       ierr = PetscMemcpy(ba+ili*bs2,uik,bs2*sizeof(MatScalar));CHKERRQ(ierr);
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++) {
634           /* rtmp += -U(i,k)^T * U_bar(i,j) */
635           rtmp_ptr = rtmp + bj[j]*bs2;
636           u = ba + j*bs2;
637           Kernel_A_gets_A_plus_Btranspose_times_C(bs,rtmp_ptr,uik,u);
638         }
639         ierr = PetscLogFlops(bslog*(jmax-jmin));CHKERRQ(ierr);
640 
641         /* ... add i to row list for next nonzero entry */
642         il[i] = jmin;             /* update il(i) in column k+1, ... mbs-1 */
643         j     = bj[jmin];
644         jl[i] = jl[j]; jl[j] = i; /* update jl */
645       }
646       i = nexti;
647     }
648 
649     /* save nonzero entries in k-th row of U ... */
650 
651     /* invert diagonal block */
652     diag = ba+k*bs2;
653     ierr = PetscMemcpy(diag,dk,bs2*sizeof(MatScalar));CHKERRQ(ierr);
654     ierr = Kernel_A_gets_inverse_A(bs,diag,pivots,work);CHKERRQ(ierr);
655 
656     jmin = bi[k]; jmax = bi[k+1];
657     if (jmin < jmax) {
658       for (j=jmin; j<jmax; j++){
659          vj = bj[j];           /* block col. index of U */
660          u   = ba + j*bs2;
661          rtmp_ptr = rtmp + vj*bs2;
662          for (k1=0; k1<bs2; k1++){
663            *u++        = *rtmp_ptr;
664            *rtmp_ptr++ = 0.0;
665          }
666       }
667 
668       /* ... add k to row list for first nonzero entry in k-th row */
669       il[k] = jmin;
670       i     = bj[jmin];
671       jl[k] = jl[i]; jl[i] = k;
672     }
673   }
674 
675   ierr = PetscFree(rtmp);CHKERRQ(ierr);
676   ierr = PetscFree(il);CHKERRQ(ierr);
677   ierr = PetscFree(dk);CHKERRQ(ierr);
678   ierr = PetscFree(pivots);CHKERRQ(ierr);
679   if (a->permute){
680     ierr = PetscFree(aa);CHKERRQ(ierr);
681   }
682 
683   ierr = ISRestoreIndices(perm,&perm_ptr);CHKERRQ(ierr);
684   C->factor       = MAT_FACTOR_CHOLESKY;
685   C->assembled    = PETSC_TRUE;
686   C->preallocated = PETSC_TRUE;
687   ierr = PetscLogFlops(1.3333*bs*bs2*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
688   PetscFunctionReturn(0);
689 }
690 
691 #undef __FUNCT__
692 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering"
693 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering(Mat A,MatFactorInfo *info,Mat *B)
694 {
695   Mat            C = *B;
696   Mat_SeqSBAIJ   *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ *)C->data;
697   PetscErrorCode ierr;
698   PetscInt       i,j,mbs=a->mbs,*bi=b->i,*bj=b->j;
699   PetscInt       *ai,*aj,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili;
700   PetscInt       bs=A->rmap.bs,bs2 = a->bs2,bslog;
701   MatScalar      *ba = b->a,*aa,*ap,*dk,*uik;
702   MatScalar      *u,*diag,*rtmp,*rtmp_ptr;
703   MatScalar      *work;
704   PetscInt       *pivots;
705 
706   PetscFunctionBegin;
707   ierr = PetscMalloc(bs2*mbs*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
708   ierr = PetscMemzero(rtmp,bs2*mbs*sizeof(MatScalar));CHKERRQ(ierr);
709   ierr = PetscMalloc(2*mbs*sizeof(PetscInt),&il);CHKERRQ(ierr);
710   jl   = il + mbs;
711   for (i=0; i<mbs; i++) {
712     jl[i] = mbs; il[0] = 0;
713   }
714   ierr = PetscMalloc((2*bs2+bs)*sizeof(MatScalar),&dk);CHKERRQ(ierr);
715   uik  = dk + bs2;
716   work = uik + bs2;
717   ierr = PetscMalloc(bs*sizeof(PetscInt),&pivots);CHKERRQ(ierr);
718 
719   ai = a->i; aj = a->j; aa = a->a;
720 
721   /* flops in while loop */
722   bslog = 2*bs*bs2;
723 
724   /* for each row k */
725   for (k = 0; k<mbs; k++){
726 
727     /*initialize k-th row with elements nonzero in row k of A */
728     jmin = ai[k]; jmax = ai[k+1];
729     ap = aa + jmin*bs2;
730     for (j = jmin; j < jmax; j++){
731       vj = aj[j];         /* block col. index */
732       rtmp_ptr = rtmp + vj*bs2;
733       for (i=0; i<bs2; i++) *rtmp_ptr++ = *ap++;
734     }
735 
736     /* modify k-th row by adding in those rows i with U(i,k) != 0 */
737     ierr = PetscMemcpy(dk,rtmp+k*bs2,bs2*sizeof(MatScalar));CHKERRQ(ierr);
738     i = jl[k]; /* first row to be added to k_th row  */
739 
740     while (i < k){
741       nexti = jl[i]; /* next row to be added to k_th row */
742 
743       /* compute multiplier */
744       ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
745 
746       /* uik = -inv(Di)*U_bar(i,k) */
747       diag = ba + i*bs2;
748       u    = ba + ili*bs2;
749       ierr = PetscMemzero(uik,bs2*sizeof(MatScalar));CHKERRQ(ierr);
750       Kernel_A_gets_A_minus_B_times_C(bs,uik,diag,u);
751 
752       /* update D(k) += -U(i,k)^T * U_bar(i,k) */
753       Kernel_A_gets_A_plus_Btranspose_times_C(bs,dk,uik,u);
754       ierr = PetscLogFlops(bslog*2);CHKERRQ(ierr);
755 
756       /* update -U(i,k) */
757       ierr = PetscMemcpy(ba+ili*bs2,uik,bs2*sizeof(MatScalar));CHKERRQ(ierr);
758 
759       /* add multiple of row i to k-th row ... */
760       jmin = ili + 1; jmax = bi[i+1];
761       if (jmin < jmax){
762         for (j=jmin; j<jmax; j++) {
763           /* rtmp += -U(i,k)^T * U_bar(i,j) */
764           rtmp_ptr = rtmp + bj[j]*bs2;
765           u = ba + j*bs2;
766           Kernel_A_gets_A_plus_Btranspose_times_C(bs,rtmp_ptr,uik,u);
767         }
768         ierr = PetscLogFlops(bslog*(jmax-jmin));CHKERRQ(ierr);
769 
770         /* ... add i to row list for next nonzero entry */
771         il[i] = jmin;             /* update il(i) in column k+1, ... mbs-1 */
772         j     = bj[jmin];
773         jl[i] = jl[j]; jl[j] = i; /* update jl */
774       }
775       i = nexti;
776     }
777 
778     /* save nonzero entries in k-th row of U ... */
779 
780     /* invert diagonal block */
781     diag = ba+k*bs2;
782     ierr = PetscMemcpy(diag,dk,bs2*sizeof(MatScalar));CHKERRQ(ierr);
783     ierr = Kernel_A_gets_inverse_A(bs,diag,pivots,work);CHKERRQ(ierr);
784 
785     jmin = bi[k]; jmax = bi[k+1];
786     if (jmin < jmax) {
787       for (j=jmin; j<jmax; j++){
788          vj = bj[j];           /* block col. index of U */
789          u   = ba + j*bs2;
790          rtmp_ptr = rtmp + vj*bs2;
791          for (k1=0; k1<bs2; k1++){
792            *u++        = *rtmp_ptr;
793            *rtmp_ptr++ = 0.0;
794          }
795       }
796 
797       /* ... add k to row list for first nonzero entry in k-th row */
798       il[k] = jmin;
799       i     = bj[jmin];
800       jl[k] = jl[i]; jl[i] = k;
801     }
802   }
803 
804   ierr = PetscFree(rtmp);CHKERRQ(ierr);
805   ierr = PetscFree(il);CHKERRQ(ierr);
806   ierr = PetscFree(dk);CHKERRQ(ierr);
807   ierr = PetscFree(pivots);CHKERRQ(ierr);
808 
809   C->factor    = MAT_FACTOR_CHOLESKY;
810   C->assembled = PETSC_TRUE;
811   C->preallocated = PETSC_TRUE;
812   ierr = PetscLogFlops(1.3333*bs*bs2*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
813   PetscFunctionReturn(0);
814 }
815 
816 /*
817     Numeric U^T*D*U factorization for SBAIJ format. Modified from SNF of YSMP.
818     Version for blocks 2 by 2.
819 */
820 #undef __FUNCT__
821 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_2"
822 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_2(Mat A,MatFactorInfo *info,Mat *B)
823 {
824   Mat            C = *B;
825   Mat_SeqSBAIJ   *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ *)C->data;
826   IS             perm = b->row;
827   PetscErrorCode ierr;
828   PetscInt       *perm_ptr,i,j,mbs=a->mbs,*bi=b->i,*bj=b->j;
829   PetscInt       *ai,*aj,*a2anew,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili;
830   MatScalar      *ba = b->a,*aa,*ap,*dk,*uik;
831   MatScalar      *u,*diag,*rtmp,*rtmp_ptr;
832   PetscReal      shift = info->shiftinblocks;
833 
834   PetscFunctionBegin;
835   /* initialization */
836   /* il and jl record the first nonzero element in each row of the accessing
837      window U(0:k, k:mbs-1).
838      jl:    list of rows to be added to uneliminated rows
839             i>= k: jl(i) is the first row to be added to row i
840             i<  k: jl(i) is the row following row i in some list of rows
841             jl(i) = mbs indicates the end of a list
842      il(i): points to the first nonzero element in columns k,...,mbs-1 of
843             row i of U */
844   ierr = PetscMalloc(4*mbs*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
845   ierr = PetscMemzero(rtmp,4*mbs*sizeof(MatScalar));CHKERRQ(ierr);
846   ierr = PetscMalloc(2*mbs*sizeof(PetscInt),&il);CHKERRQ(ierr);
847   jl   = il + mbs;
848   for (i=0; i<mbs; i++) {
849     jl[i] = mbs; il[0] = 0;
850   }
851   ierr = PetscMalloc(8*sizeof(MatScalar),&dk);CHKERRQ(ierr);
852   uik  = dk + 4;
853   ierr = ISGetIndices(perm,&perm_ptr);CHKERRQ(ierr);
854 
855   /* check permutation */
856   if (!a->permute){
857     ai = a->i; aj = a->j; aa = a->a;
858   } else {
859     ai   = a->inew; aj = a->jnew;
860     ierr = PetscMalloc(4*ai[mbs]*sizeof(MatScalar),&aa);CHKERRQ(ierr);
861     ierr = PetscMemcpy(aa,a->a,4*ai[mbs]*sizeof(MatScalar));CHKERRQ(ierr);
862     ierr = PetscMalloc(ai[mbs]*sizeof(PetscInt),&a2anew);CHKERRQ(ierr);
863     ierr = PetscMemcpy(a2anew,a->a2anew,(ai[mbs])*sizeof(PetscInt));CHKERRQ(ierr);
864 
865     for (i=0; i<mbs; i++){
866       jmin = ai[i]; jmax = ai[i+1];
867       for (j=jmin; j<jmax; j++){
868         while (a2anew[j] != j){
869           k = a2anew[j]; a2anew[j] = a2anew[k]; a2anew[k] = k;
870           for (k1=0; k1<4; k1++){
871             dk[k1]       = aa[k*4+k1];
872             aa[k*4+k1] = aa[j*4+k1];
873             aa[j*4+k1] = dk[k1];
874           }
875         }
876         /* transform columnoriented blocks that lie in the lower triangle to roworiented blocks */
877         if (i > aj[j]){
878           /* printf("change orientation, row: %d, col: %d\n",i,aj[j]); */
879           ap = aa + j*4;     /* ptr to the beginning of the block */
880           dk[1] = ap[1];     /* swap ap[1] and ap[2] */
881           ap[1] = ap[2];
882           ap[2] = dk[1];
883         }
884       }
885     }
886     ierr = PetscFree(a2anew);CHKERRQ(ierr);
887   }
888 
889   /* for each row k */
890   for (k = 0; k<mbs; k++){
891 
892     /*initialize k-th row with elements nonzero in row perm(k) of A */
893     jmin = ai[perm_ptr[k]]; jmax = ai[perm_ptr[k]+1];
894     ap = aa + jmin*4;
895     for (j = jmin; j < jmax; j++){
896       vj = perm_ptr[aj[j]];         /* block col. index */
897       rtmp_ptr = rtmp + vj*4;
898       for (i=0; i<4; i++) *rtmp_ptr++ = *ap++;
899     }
900 
901     /* modify k-th row by adding in those rows i with U(i,k) != 0 */
902     ierr = PetscMemcpy(dk,rtmp+k*4,4*sizeof(MatScalar));CHKERRQ(ierr);
903     i = jl[k]; /* first row to be added to k_th row  */
904 
905     while (i < k){
906       nexti = jl[i]; /* next row to be added to k_th row */
907 
908       /* compute multiplier */
909       ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
910 
911       /* uik = -inv(Di)*U_bar(i,k): - ba[ili]*ba[i] */
912       diag = ba + i*4;
913       u    = ba + ili*4;
914       uik[0] = -(diag[0]*u[0] + diag[2]*u[1]);
915       uik[1] = -(diag[1]*u[0] + diag[3]*u[1]);
916       uik[2] = -(diag[0]*u[2] + diag[2]*u[3]);
917       uik[3] = -(diag[1]*u[2] + diag[3]*u[3]);
918 
919       /* update D(k) += -U(i,k)^T * U_bar(i,k): dk += uik*ba[ili] */
920       dk[0] += uik[0]*u[0] + uik[1]*u[1];
921       dk[1] += uik[2]*u[0] + uik[3]*u[1];
922       dk[2] += uik[0]*u[2] + uik[1]*u[3];
923       dk[3] += uik[2]*u[2] + uik[3]*u[3];
924 
925       ierr = PetscLogFlops(16*2);CHKERRQ(ierr);
926 
927       /* update -U(i,k): ba[ili] = uik */
928       ierr = PetscMemcpy(ba+ili*4,uik,4*sizeof(MatScalar));CHKERRQ(ierr);
929 
930       /* add multiple of row i to k-th row ... */
931       jmin = ili + 1; jmax = bi[i+1];
932       if (jmin < jmax){
933         for (j=jmin; j<jmax; j++) {
934           /* rtmp += -U(i,k)^T * U_bar(i,j): rtmp[bj[j]] += uik*ba[j]; */
935           rtmp_ptr = rtmp + bj[j]*4;
936           u = ba + j*4;
937           rtmp_ptr[0] += uik[0]*u[0] + uik[1]*u[1];
938           rtmp_ptr[1] += uik[2]*u[0] + uik[3]*u[1];
939           rtmp_ptr[2] += uik[0]*u[2] + uik[1]*u[3];
940           rtmp_ptr[3] += uik[2]*u[2] + uik[3]*u[3];
941         }
942         ierr = PetscLogFlops(16*(jmax-jmin));CHKERRQ(ierr);
943 
944         /* ... add i to row list for next nonzero entry */
945         il[i] = jmin;             /* update il(i) in column k+1, ... mbs-1 */
946         j     = bj[jmin];
947         jl[i] = jl[j]; jl[j] = i; /* update jl */
948       }
949       i = nexti;
950     }
951 
952     /* save nonzero entries in k-th row of U ... */
953 
954     /* invert diagonal block */
955     diag = ba+k*4;
956     ierr = PetscMemcpy(diag,dk,4*sizeof(MatScalar));CHKERRQ(ierr);
957     ierr = Kernel_A_gets_inverse_A_2(diag,shift);CHKERRQ(ierr);
958 
959     jmin = bi[k]; jmax = bi[k+1];
960     if (jmin < jmax) {
961       for (j=jmin; j<jmax; j++){
962          vj = bj[j];           /* block col. index of U */
963          u   = ba + j*4;
964          rtmp_ptr = rtmp + vj*4;
965          for (k1=0; k1<4; k1++){
966            *u++        = *rtmp_ptr;
967            *rtmp_ptr++ = 0.0;
968          }
969       }
970 
971       /* ... add k to row list for first nonzero entry in k-th row */
972       il[k] = jmin;
973       i     = bj[jmin];
974       jl[k] = jl[i]; jl[i] = k;
975     }
976   }
977 
978   ierr = PetscFree(rtmp);CHKERRQ(ierr);
979   ierr = PetscFree(il);CHKERRQ(ierr);
980   ierr = PetscFree(dk);CHKERRQ(ierr);
981   if (a->permute) {
982     ierr = PetscFree(aa);CHKERRQ(ierr);
983   }
984   ierr = ISRestoreIndices(perm,&perm_ptr);CHKERRQ(ierr);
985   C->factor    = MAT_FACTOR_CHOLESKY;
986   C->assembled = PETSC_TRUE;
987   C->preallocated = PETSC_TRUE;
988   ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
989   PetscFunctionReturn(0);
990 }
991 
992 /*
993       Version for when blocks are 2 by 2 Using natural ordering
994 */
995 #undef __FUNCT__
996 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering"
997 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering(Mat A,MatFactorInfo *info,Mat *B)
998 {
999   Mat            C = *B;
1000   Mat_SeqSBAIJ   *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ *)C->data;
1001   PetscErrorCode ierr;
1002   PetscInt       i,j,mbs=a->mbs,*bi=b->i,*bj=b->j;
1003   PetscInt       *ai,*aj,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili;
1004   MatScalar      *ba = b->a,*aa,*ap,*dk,*uik;
1005   MatScalar      *u,*diag,*rtmp,*rtmp_ptr;
1006   PetscReal      shift = info->shiftinblocks;
1007 
1008   PetscFunctionBegin;
1009   /* initialization */
1010   /* il and jl record the first nonzero element in each row of the accessing
1011      window U(0:k, k:mbs-1).
1012      jl:    list of rows to be added to uneliminated rows
1013             i>= k: jl(i) is the first row to be added to row i
1014             i<  k: jl(i) is the row following row i in some list of rows
1015             jl(i) = mbs indicates the end of a list
1016      il(i): points to the first nonzero element in columns k,...,mbs-1 of
1017             row i of U */
1018   ierr = PetscMalloc(4*mbs*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
1019   ierr = PetscMemzero(rtmp,4*mbs*sizeof(MatScalar));CHKERRQ(ierr);
1020   ierr = PetscMalloc(2*mbs*sizeof(PetscInt),&il);CHKERRQ(ierr);
1021   jl   = il + mbs;
1022   for (i=0; i<mbs; i++) {
1023     jl[i] = mbs; il[0] = 0;
1024   }
1025   ierr = PetscMalloc(8*sizeof(MatScalar),&dk);CHKERRQ(ierr);
1026   uik  = dk + 4;
1027 
1028   ai = a->i; aj = a->j; aa = a->a;
1029 
1030   /* for each row k */
1031   for (k = 0; k<mbs; k++){
1032 
1033     /*initialize k-th row with elements nonzero in row k of A */
1034     jmin = ai[k]; jmax = ai[k+1];
1035     ap = aa + jmin*4;
1036     for (j = jmin; j < jmax; j++){
1037       vj = aj[j];         /* block col. index */
1038       rtmp_ptr = rtmp + vj*4;
1039       for (i=0; i<4; i++) *rtmp_ptr++ = *ap++;
1040     }
1041 
1042     /* modify k-th row by adding in those rows i with U(i,k) != 0 */
1043     ierr = PetscMemcpy(dk,rtmp+k*4,4*sizeof(MatScalar));CHKERRQ(ierr);
1044     i = jl[k]; /* first row to be added to k_th row  */
1045 
1046     while (i < k){
1047       nexti = jl[i]; /* next row to be added to k_th row */
1048 
1049       /* compute multiplier */
1050       ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
1051 
1052       /* uik = -inv(Di)*U_bar(i,k): - ba[ili]*ba[i] */
1053       diag = ba + i*4;
1054       u    = ba + ili*4;
1055       uik[0] = -(diag[0]*u[0] + diag[2]*u[1]);
1056       uik[1] = -(diag[1]*u[0] + diag[3]*u[1]);
1057       uik[2] = -(diag[0]*u[2] + diag[2]*u[3]);
1058       uik[3] = -(diag[1]*u[2] + diag[3]*u[3]);
1059 
1060       /* update D(k) += -U(i,k)^T * U_bar(i,k): dk += uik*ba[ili] */
1061       dk[0] += uik[0]*u[0] + uik[1]*u[1];
1062       dk[1] += uik[2]*u[0] + uik[3]*u[1];
1063       dk[2] += uik[0]*u[2] + uik[1]*u[3];
1064       dk[3] += uik[2]*u[2] + uik[3]*u[3];
1065 
1066       ierr = PetscLogFlops(16*2);CHKERRQ(ierr);
1067 
1068       /* update -U(i,k): ba[ili] = uik */
1069       ierr = PetscMemcpy(ba+ili*4,uik,4*sizeof(MatScalar));CHKERRQ(ierr);
1070 
1071       /* add multiple of row i to k-th row ... */
1072       jmin = ili + 1; jmax = bi[i+1];
1073       if (jmin < jmax){
1074         for (j=jmin; j<jmax; j++) {
1075           /* rtmp += -U(i,k)^T * U_bar(i,j): rtmp[bj[j]] += uik*ba[j]; */
1076           rtmp_ptr = rtmp + bj[j]*4;
1077           u = ba + j*4;
1078           rtmp_ptr[0] += uik[0]*u[0] + uik[1]*u[1];
1079           rtmp_ptr[1] += uik[2]*u[0] + uik[3]*u[1];
1080           rtmp_ptr[2] += uik[0]*u[2] + uik[1]*u[3];
1081           rtmp_ptr[3] += uik[2]*u[2] + uik[3]*u[3];
1082         }
1083         ierr = PetscLogFlops(16*(jmax-jmin));CHKERRQ(ierr);
1084 
1085         /* ... add i to row list for next nonzero entry */
1086         il[i] = jmin;             /* update il(i) in column k+1, ... mbs-1 */
1087         j     = bj[jmin];
1088         jl[i] = jl[j]; jl[j] = i; /* update jl */
1089       }
1090       i = nexti;
1091     }
1092 
1093     /* save nonzero entries in k-th row of U ... */
1094 
1095     /* invert diagonal block */
1096     diag = ba+k*4;
1097     ierr = PetscMemcpy(diag,dk,4*sizeof(MatScalar));CHKERRQ(ierr);
1098     ierr = Kernel_A_gets_inverse_A_2(diag,shift);CHKERRQ(ierr);
1099 
1100     jmin = bi[k]; jmax = bi[k+1];
1101     if (jmin < jmax) {
1102       for (j=jmin; j<jmax; j++){
1103          vj = bj[j];           /* block col. index of U */
1104          u   = ba + j*4;
1105          rtmp_ptr = rtmp + vj*4;
1106          for (k1=0; k1<4; k1++){
1107            *u++        = *rtmp_ptr;
1108            *rtmp_ptr++ = 0.0;
1109          }
1110       }
1111 
1112       /* ... add k to row list for first nonzero entry in k-th row */
1113       il[k] = jmin;
1114       i     = bj[jmin];
1115       jl[k] = jl[i]; jl[i] = k;
1116     }
1117   }
1118 
1119   ierr = PetscFree(rtmp);CHKERRQ(ierr);
1120   ierr = PetscFree(il);CHKERRQ(ierr);
1121   ierr = PetscFree(dk);CHKERRQ(ierr);
1122 
1123   C->factor    = MAT_FACTOR_CHOLESKY;
1124   C->assembled = PETSC_TRUE;
1125   C->preallocated = PETSC_TRUE;
1126   ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
1127   PetscFunctionReturn(0);
1128 }
1129 
1130 /*
1131     Numeric U^T*D*U factorization for SBAIJ format.
1132     Version for blocks are 1 by 1.
1133 */
1134 #undef __FUNCT__
1135 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_1"
1136 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_1(Mat A,MatFactorInfo *info,Mat *B)
1137 {
1138   Mat            C = *B;
1139   Mat_SeqSBAIJ   *a=(Mat_SeqSBAIJ*)A->data,*b=(Mat_SeqSBAIJ *)C->data;
1140   IS             ip=b->row;
1141   PetscErrorCode ierr;
1142   PetscInt       *rip,i,j,mbs=a->mbs,*bi=b->i,*bj=b->j,*bcol;
1143   PetscInt       *ai,*aj,*a2anew;
1144   PetscInt       k,jmin,jmax,*jl,*il,col,nexti,ili,nz;
1145   MatScalar      *rtmp,*ba=b->a,*bval,*aa,dk,uikdi;
1146   PetscReal      zeropivot,rs,shiftnz;
1147   PetscReal      shiftpd;
1148   ChShift_Ctx    sctx;
1149   PetscInt       newshift;
1150 
1151   PetscFunctionBegin;
1152   /* initialization */
1153   shiftnz   = info->shiftnz;
1154   shiftpd   = info->shiftpd;
1155   zeropivot = info->zeropivot;
1156 
1157   ierr  = ISGetIndices(ip,&rip);CHKERRQ(ierr);
1158   if (!a->permute){
1159     ai = a->i; aj = a->j; aa = a->a;
1160   } else {
1161     ai = a->inew; aj = a->jnew;
1162     nz = ai[mbs];
1163     ierr = PetscMalloc(nz*sizeof(MatScalar),&aa);CHKERRQ(ierr);
1164     a2anew = a->a2anew;
1165     bval   = a->a;
1166     for (j=0; j<nz; j++){
1167       aa[a2anew[j]] = *(bval++);
1168     }
1169   }
1170 
1171   /* initialization */
1172   /* il and jl record the first nonzero element in each row of the accessing
1173      window U(0:k, k:mbs-1).
1174      jl:    list of rows to be added to uneliminated rows
1175             i>= k: jl(i) is the first row to be added to row i
1176             i<  k: jl(i) is the row following row i in some list of rows
1177             jl(i) = mbs indicates the end of a list
1178      il(i): points to the first nonzero element in columns k,...,mbs-1 of
1179             row i of U */
1180   nz   = (2*mbs+1)*sizeof(PetscInt)+mbs*sizeof(MatScalar);
1181   ierr = PetscMalloc(nz,&il);CHKERRQ(ierr);
1182   jl   = il + mbs;
1183   rtmp = (MatScalar*)(jl + mbs);
1184 
1185   sctx.shift_amount = 0;
1186   sctx.nshift       = 0;
1187   do {
1188     sctx.chshift = PETSC_FALSE;
1189     for (i=0; i<mbs; i++) {
1190       rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0;
1191     }
1192 
1193     for (k = 0; k<mbs; k++){
1194       /*initialize k-th row by the perm[k]-th row of A */
1195       jmin = ai[rip[k]]; jmax = ai[rip[k]+1];
1196       bval = ba + bi[k];
1197       for (j = jmin; j < jmax; j++){
1198         col = rip[aj[j]];
1199         rtmp[col] = aa[j];
1200         *bval++  = 0.0; /* for in-place factorization */
1201       }
1202 
1203       /* shift the diagonal of the matrix */
1204       if (sctx.nshift) rtmp[k] += sctx.shift_amount;
1205 
1206       /* modify k-th row by adding in those rows i with U(i,k)!=0 */
1207       dk = rtmp[k];
1208       i = jl[k]; /* first row to be added to k_th row  */
1209 
1210       while (i < k){
1211         nexti = jl[i]; /* next row to be added to k_th row */
1212 
1213         /* compute multiplier, update diag(k) and U(i,k) */
1214         ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
1215         uikdi = - ba[ili]*ba[bi[i]];  /* diagonal(k) */
1216         dk += uikdi*ba[ili];
1217         ba[ili] = uikdi; /* -U(i,k) */
1218 
1219         /* add multiple of row i to k-th row */
1220         jmin = ili + 1; jmax = bi[i+1];
1221         if (jmin < jmax){
1222           for (j=jmin; j<jmax; j++) rtmp[bj[j]] += uikdi*ba[j];
1223           ierr = PetscLogFlops(2*(jmax-jmin));CHKERRQ(ierr);
1224 
1225           /* update il and jl for row i */
1226           il[i] = jmin;
1227           j = bj[jmin]; jl[i] = jl[j]; jl[j] = i;
1228         }
1229         i = nexti;
1230       }
1231 
1232       /* shift the diagonals when zero pivot is detected */
1233       /* compute rs=sum of abs(off-diagonal) */
1234       rs   = 0.0;
1235       jmin = bi[k]+1;
1236       nz   = bi[k+1] - jmin;
1237       if (nz){
1238         bcol = bj + jmin;
1239         while (nz--){
1240           rs += PetscAbsScalar(rtmp[*bcol]);
1241           bcol++;
1242         }
1243       }
1244 
1245       sctx.rs = rs;
1246       sctx.pv = dk;
1247       ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr);
1248       if (newshift == 1) break;    /* sctx.shift_amount is updated */
1249 
1250       /* copy data into U(k,:) */
1251       ba[bi[k]] = 1.0/dk; /* U(k,k) */
1252       jmin = bi[k]+1; jmax = bi[k+1];
1253       if (jmin < jmax) {
1254         for (j=jmin; j<jmax; j++){
1255           col = bj[j]; ba[j] = rtmp[col]; rtmp[col] = 0.0;
1256         }
1257         /* add the k-th row into il and jl */
1258         il[k] = jmin;
1259         i = bj[jmin]; jl[k] = jl[i]; jl[i] = k;
1260       }
1261     }
1262   } while (sctx.chshift);
1263   ierr = PetscFree(il);CHKERRQ(ierr);
1264   if (a->permute){ierr = PetscFree(aa);CHKERRQ(ierr);}
1265 
1266   ierr = ISRestoreIndices(ip,&rip);CHKERRQ(ierr);
1267   C->factor       = MAT_FACTOR_CHOLESKY;
1268   C->assembled    = PETSC_TRUE;
1269   C->preallocated = PETSC_TRUE;
1270   ierr = PetscLogFlops(C->rmap.N);CHKERRQ(ierr);
1271   if (sctx.nshift){
1272     if (shiftnz) {
1273       ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
1274     } else if (shiftpd) {
1275       ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
1276     }
1277   }
1278   PetscFunctionReturn(0);
1279 }
1280 
1281 /*
1282   Version for when blocks are 1 by 1 Using natural ordering
1283 */
1284 #undef __FUNCT__
1285 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering"
1286 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering(Mat A,MatFactorInfo *info,Mat *B)
1287 {
1288   Mat            C = *B;
1289   Mat_SeqSBAIJ   *a=(Mat_SeqSBAIJ*)A->data,*b=(Mat_SeqSBAIJ *)C->data;
1290   PetscErrorCode ierr;
1291   PetscInt       i,j,mbs = a->mbs;
1292   PetscInt       *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
1293   PetscInt       k,jmin,*jl,*il,nexti,ili,*acol,*bcol,nz;
1294   MatScalar      *rtmp,*ba=b->a,*aa=a->a,dk,uikdi,*aval,*bval;
1295   PetscReal      zeropivot,rs,shiftnz;
1296   PetscReal      shiftpd;
1297   ChShift_Ctx    sctx;
1298   PetscInt       newshift;
1299 
1300   PetscFunctionBegin;
1301   /* initialization */
1302   shiftnz   = info->shiftnz;
1303   shiftpd   = info->shiftpd;
1304   zeropivot = info->zeropivot;
1305 
1306   /* il and jl record the first nonzero element in each row of the accessing
1307      window U(0:k, k:mbs-1).
1308      jl:    list of rows to be added to uneliminated rows
1309             i>= k: jl(i) is the first row to be added to row i
1310             i<  k: jl(i) is the row following row i in some list of rows
1311             jl(i) = mbs indicates the end of a list
1312      il(i): points to the first nonzero element in U(i,k:mbs-1)
1313   */
1314   ierr = PetscMalloc(mbs*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
1315   ierr = PetscMalloc(2*mbs*sizeof(PetscInt),&il);CHKERRQ(ierr);
1316   jl   = il + mbs;
1317 
1318   sctx.shift_amount = 0;
1319   sctx.nshift       = 0;
1320   do {
1321     sctx.chshift = PETSC_FALSE;
1322     for (i=0; i<mbs; i++) {
1323       rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0;
1324     }
1325 
1326     for (k = 0; k<mbs; k++){
1327       /*initialize k-th row with elements nonzero in row perm(k) of A */
1328       nz   = ai[k+1] - ai[k];
1329       acol = aj + ai[k];
1330       aval = aa + ai[k];
1331       bval = ba + bi[k];
1332       while (nz -- ){
1333         rtmp[*acol++] = *aval++;
1334         *bval++       = 0.0; /* for in-place factorization */
1335       }
1336 
1337       /* shift the diagonal of the matrix */
1338       if (sctx.nshift) rtmp[k] += sctx.shift_amount;
1339 
1340       /* modify k-th row by adding in those rows i with U(i,k)!=0 */
1341       dk = rtmp[k];
1342       i  = jl[k]; /* first row to be added to k_th row  */
1343 
1344       while (i < k){
1345         nexti = jl[i]; /* next row to be added to k_th row */
1346         /* compute multiplier, update D(k) and U(i,k) */
1347         ili   = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
1348         uikdi = - ba[ili]*ba[bi[i]];
1349         dk   += uikdi*ba[ili];
1350         ba[ili] = uikdi; /* -U(i,k) */
1351 
1352         /* add multiple of row i to k-th row ... */
1353         jmin = ili + 1;
1354         nz   = bi[i+1] - jmin;
1355         if (nz > 0){
1356           bcol = bj + jmin;
1357           bval = ba + jmin;
1358           ierr = PetscLogFlops(2*nz);CHKERRQ(ierr);
1359           while (nz --) rtmp[*bcol++] += uikdi*(*bval++);
1360 
1361           /* update il and jl for i-th row */
1362           il[i] = jmin;
1363           j = bj[jmin]; jl[i] = jl[j]; jl[j] = i;
1364         }
1365         i = nexti;
1366       }
1367 
1368       /* shift the diagonals when zero pivot is detected */
1369       /* compute rs=sum of abs(off-diagonal) */
1370       rs   = 0.0;
1371       jmin = bi[k]+1;
1372       nz   = bi[k+1] - jmin;
1373       if (nz){
1374         bcol = bj + jmin;
1375         while (nz--){
1376           rs += PetscAbsScalar(rtmp[*bcol]);
1377           bcol++;
1378         }
1379       }
1380 
1381       sctx.rs = rs;
1382       sctx.pv = dk;
1383       ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr);
1384       if (newshift == 1) break;    /* sctx.shift_amount is updated */
1385 
1386       /* copy data into U(k,:) */
1387       ba[bi[k]] = 1.0/dk;
1388       jmin      = bi[k]+1;
1389       nz        = bi[k+1] - jmin;
1390       if (nz){
1391         bcol = bj + jmin;
1392         bval = ba + jmin;
1393         while (nz--){
1394           *bval++       = rtmp[*bcol];
1395           rtmp[*bcol++] = 0.0;
1396         }
1397         /* add k-th row into il and jl */
1398         il[k] = jmin;
1399         i = bj[jmin]; jl[k] = jl[i]; jl[i] = k;
1400       }
1401     } /* end of for (k = 0; k<mbs; k++) */
1402   } while (sctx.chshift);
1403   ierr = PetscFree(rtmp);CHKERRQ(ierr);
1404   ierr = PetscFree(il);CHKERRQ(ierr);
1405 
1406   C->factor       = MAT_FACTOR_CHOLESKY;
1407   C->assembled    = PETSC_TRUE;
1408   C->preallocated = PETSC_TRUE;
1409   ierr = PetscLogFlops(C->rmap.N);CHKERRQ(ierr);
1410   if (sctx.nshift){
1411     if (shiftnz) {
1412       ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
1413     } else if (shiftpd) {
1414       ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
1415     }
1416   }
1417   PetscFunctionReturn(0);
1418 }
1419 
1420 #undef __FUNCT__
1421 #define __FUNCT__ "MatCholeskyFactor_SeqSBAIJ"
1422 PetscErrorCode MatCholeskyFactor_SeqSBAIJ(Mat A,IS perm,MatFactorInfo *info)
1423 {
1424   PetscErrorCode ierr;
1425   Mat            C;
1426 
1427   PetscFunctionBegin;
1428   ierr = MatCholeskyFactorSymbolic(A,perm,info,&C);CHKERRQ(ierr);
1429   ierr = MatCholeskyFactorNumeric(A,info,&C);CHKERRQ(ierr);
1430   ierr = MatHeaderCopy(A,C);CHKERRQ(ierr);
1431   PetscFunctionReturn(0);
1432 }
1433 
1434 
1435