Actual source code: sbaijov.c
petsc-3.6.1 2015-08-06
2: /*
3: Routines to compute overlapping regions of a parallel MPI matrix.
4: Used for finding submatrices that were shared across processors.
5: */
6: #include <../src/mat/impls/sbaij/mpi/mpisbaij.h>
7: #include <petscbt.h>
9: static PetscErrorCode MatIncreaseOverlap_MPISBAIJ_Once(Mat,PetscInt,IS*);
10: static PetscErrorCode MatIncreaseOverlap_MPISBAIJ_Local(Mat,PetscInt*,PetscInt,PetscInt*,PetscBT*);
14: PetscErrorCode MatIncreaseOverlap_MPISBAIJ(Mat C,PetscInt is_max,IS is[],PetscInt ov)
15: {
17: PetscInt i,N=C->cmap->N, bs=C->rmap->bs,M=C->rmap->N,Mbs=M/bs,*nidx,isz,iov;
18: IS *is_new,*is_row;
19: Mat *submats;
20: Mat_MPISBAIJ *c=(Mat_MPISBAIJ*)C->data;
21: Mat_SeqSBAIJ *asub_i;
22: PetscBT table;
23: PetscInt *ai,brow,nz,nis,l,nmax,nstages_local,nstages,max_no,pos;
24: const PetscInt *idx;
25: PetscBool flg,*allcolumns,*allrows;
28: PetscMalloc1(is_max,&is_new);
29: /* Convert the indices into block format */
30: ISCompressIndicesGeneral(N,C->rmap->n,bs,is_max,is,is_new);
31: if (ov < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative overlap specified\n");
33: /* ----- previous non-scalable implementation ----- */
34: flg = PETSC_FALSE;
35: PetscOptionsHasName(NULL, "-IncreaseOverlap_old", &flg);
36: if (flg) { /* previous non-scalable implementation */
37: printf("use previous non-scalable implementation...\n");
38: for (i=0; i<ov; ++i) {
39: MatIncreaseOverlap_MPISBAIJ_Once(C,is_max,is_new);
40: }
41: } else { /* implementation using modified BAIJ routines */
43: PetscMalloc1(Mbs+1,&nidx);
44: PetscBTCreate(Mbs,&table); /* for column search */
45: PetscMalloc2(is_max+1,&allcolumns,is_max+1,&allrows);
47: /* Create is_row */
48: PetscMalloc1(is_max,&is_row);
49: ISCreateStride(PETSC_COMM_SELF,Mbs,0,1,&is_row[0]);
51: allrows[0] = PETSC_TRUE;
52: for (i=1; i<is_max; i++) {
53: is_row[i] = is_row[0]; /* reuse is_row[0] */
54: allrows[i] = PETSC_TRUE;
55: }
57: /* Allocate memory to hold all the submatrices - Modified from MatGetSubMatrices_MPIBAIJ() */
58: PetscMalloc1(is_max+1,&submats);
60: /* Check for special case: each processor gets entire matrix columns */
61: for (i=0; i<is_max; i++) {
62: ISIdentity(is_new[i],&flg);
63: ISGetLocalSize(is_new[i],&isz);
64: if (flg && isz == Mbs) {
65: allcolumns[i] = PETSC_TRUE;
66: } else {
67: allcolumns[i] = PETSC_FALSE;
68: }
69: }
71: /* Determine the number of stages through which submatrices are done */
72: nmax = 20*1000000 / (c->Nbs * sizeof(PetscInt));
73: if (!nmax) nmax = 1;
74: nstages_local = is_max/nmax + ((is_max % nmax) ? 1 : 0);
76: /* Make sure every processor loops through the nstages */
77: MPI_Allreduce(&nstages_local,&nstages,1,MPIU_INT,MPI_MAX,PetscObjectComm((PetscObject)C));
79: for (iov=0; iov<ov; ++iov) {
80: /* 1) Get submats for column search */
81: for (i=0,pos=0; i<nstages; i++) {
82: if (pos+nmax <= is_max) max_no = nmax;
83: else if (pos == is_max) max_no = 0;
84: else max_no = is_max-pos;
85: c->ijonly = PETSC_TRUE;
86: MatGetSubMatrices_MPIBAIJ_local(C,max_no,is_row+pos,is_new+pos,MAT_INITIAL_MATRIX,allrows+pos,allcolumns+pos,submats+pos);
87: pos += max_no;
88: }
90: /* 2) Row search */
91: MatIncreaseOverlap_MPIBAIJ_Once(C,is_max,is_new);
93: /* 3) Column search */
94: for (i=0; i<is_max; i++) {
95: asub_i = (Mat_SeqSBAIJ*)submats[i]->data;
96: ai = asub_i->i;;
98: /* put is_new obtained from MatIncreaseOverlap_MPIBAIJ() to table */
99: PetscBTMemzero(Mbs,table);
101: ISGetIndices(is_new[i],&idx);
102: ISGetLocalSize(is_new[i],&nis);
103: for (l=0; l<nis; l++) {
104: PetscBTSet(table,idx[l]);
105: nidx[l] = idx[l];
106: }
107: isz = nis;
109: /* add column entries to table */
110: for (brow=0; brow<Mbs; brow++) {
111: nz = ai[brow+1] - ai[brow];
112: if (nz) {
113: if (!PetscBTLookupSet(table,brow)) nidx[isz++] = brow;
114: }
115: }
116: ISRestoreIndices(is_new[i],&idx);
117: ISDestroy(&is_new[i]);
119: /* create updated is_new */
120: ISCreateGeneral(PETSC_COMM_SELF,isz,nidx,PETSC_COPY_VALUES,is_new+i);
121: }
123: /* Free tmp spaces */
124: for (i=0; i<is_max; i++) {
125: MatDestroy(&submats[i]);
126: }
127: }
128: PetscFree2(allcolumns,allrows);
129: PetscBTDestroy(&table);
130: PetscFree(submats);
131: ISDestroy(&is_row[0]);
132: PetscFree(is_row);
133: PetscFree(nidx);
135: }
137: for (i=0; i<is_max; i++) {ISDestroy(&is[i]);}
138: ISExpandIndicesGeneral(N,N,bs,is_max,is_new,is);
140: for (i=0; i<is_max; i++) {ISDestroy(&is_new[i]);}
141: PetscFree(is_new);
142: return(0);
143: }
145: typedef enum {MINE,OTHER} WhoseOwner;
146: /* data1, odata1 and odata2 are packed in the format (for communication):
147: data[0] = is_max, no of is
148: data[1] = size of is[0]
149: ...
150: data[is_max] = size of is[is_max-1]
151: data[is_max + 1] = data(is[0])
152: ...
153: data[is_max+1+sum(size of is[k]), k=0,...,i-1] = data(is[i])
154: ...
155: data2 is packed in the format (for creating output is[]):
156: data[0] = is_max, no of is
157: data[1] = size of is[0]
158: ...
159: data[is_max] = size of is[is_max-1]
160: data[is_max + 1] = data(is[0])
161: ...
162: data[is_max + 1 + Mbs*i) = data(is[i])
163: ...
164: */
167: static PetscErrorCode MatIncreaseOverlap_MPISBAIJ_Once(Mat C,PetscInt is_max,IS is[])
168: {
169: Mat_MPISBAIJ *c = (Mat_MPISBAIJ*)C->data;
171: PetscMPIInt size,rank,tag1,tag2,*len_s,nrqr,nrqs,*id_r1,*len_r1,flag,len,*iwork;
172: const PetscInt *idx_i;
173: PetscInt idx,isz,col,*n,*data1,**data1_start,*data2,*data2_i,*data,*data_i;
174: PetscInt Mbs,i,j,k,*odata1,*odata2;
175: PetscInt proc_id,**odata2_ptr,*ctable=0,*btable,len_max,len_est;
176: PetscInt proc_end=0,len_unused,nodata2;
177: PetscInt ois_max; /* max no of is[] in each of processor */
178: char *t_p;
179: MPI_Comm comm;
180: MPI_Request *s_waits1,*s_waits2,r_req;
181: MPI_Status *s_status,r_status;
182: PetscBT *table; /* mark indices of this processor's is[] */
183: PetscBT table_i;
184: PetscBT otable; /* mark indices of other processors' is[] */
185: PetscInt bs=C->rmap->bs,Bn = c->B->cmap->n,Bnbs = Bn/bs,*Bowners;
186: IS garray_local,garray_gl;
189: PetscObjectGetComm((PetscObject)C,&comm);
190: size = c->size;
191: rank = c->rank;
192: Mbs = c->Mbs;
194: PetscObjectGetNewTag((PetscObject)C,&tag1);
195: PetscObjectGetNewTag((PetscObject)C,&tag2);
197: /* create tables used in
198: step 1: table[i] - mark c->garray of proc [i]
199: step 3: table[i] - mark indices of is[i] when whose=MINE
200: table[0] - mark incideces of is[] when whose=OTHER */
201: len = PetscMax(is_max, size);
202: PetscMalloc2(len,&table,(Mbs/PETSC_BITS_PER_BYTE+1)*len,&t_p);
203: for (i=0; i<len; i++) {
204: table[i] = t_p + (Mbs/PETSC_BITS_PER_BYTE+1)*i;
205: }
207: MPI_Allreduce(&is_max,&ois_max,1,MPIU_INT,MPI_MAX,comm);
209: /* 1. Send this processor's is[] to other processors */
210: /*---------------------------------------------------*/
211: /* allocate spaces */
212: PetscMalloc1(is_max,&n);
213: len = 0;
214: for (i=0; i<is_max; i++) {
215: ISGetLocalSize(is[i],&n[i]);
216: len += n[i];
217: }
218: if (!len) {
219: is_max = 0;
220: } else {
221: len += 1 + is_max; /* max length of data1 for one processor */
222: }
225: PetscMalloc1(size*len+1,&data1);
226: PetscMalloc1(size,&data1_start);
227: for (i=0; i<size; i++) data1_start[i] = data1 + i*len;
229: PetscMalloc4(size,&len_s,size,&btable,size,&iwork,size+1,&Bowners);
231: /* gather c->garray from all processors */
232: ISCreateGeneral(comm,Bnbs,c->garray,PETSC_COPY_VALUES,&garray_local);
233: ISAllGather(garray_local, &garray_gl);
234: ISDestroy(&garray_local);
235: MPI_Allgather(&Bnbs,1,MPIU_INT,Bowners+1,1,MPIU_INT,comm);
237: Bowners[0] = 0;
238: for (i=0; i<size; i++) Bowners[i+1] += Bowners[i];
240: if (is_max) {
241: /* hash table ctable which maps c->row to proc_id) */
242: PetscMalloc1(Mbs,&ctable);
243: for (proc_id=0,j=0; proc_id<size; proc_id++) {
244: for (; j<C->rmap->range[proc_id+1]/bs; j++) ctable[j] = proc_id;
245: }
247: /* hash tables marking c->garray */
248: ISGetIndices(garray_gl,&idx_i);
249: for (i=0; i<size; i++) {
250: table_i = table[i];
251: PetscBTMemzero(Mbs,table_i);
252: for (j = Bowners[i]; j<Bowners[i+1]; j++) { /* go through B cols of proc[i]*/
253: PetscBTSet(table_i,idx_i[j]);
254: }
255: }
256: ISRestoreIndices(garray_gl,&idx_i);
257: } /* if (is_max) */
258: ISDestroy(&garray_gl);
260: /* evaluate communication - mesg to who, length, and buffer space */
261: for (i=0; i<size; i++) len_s[i] = 0;
263: /* header of data1 */
264: for (proc_id=0; proc_id<size; proc_id++) {
265: iwork[proc_id] = 0;
266: *data1_start[proc_id] = is_max;
267: data1_start[proc_id]++;
268: for (j=0; j<is_max; j++) {
269: if (proc_id == rank) {
270: *data1_start[proc_id] = n[j];
271: } else {
272: *data1_start[proc_id] = 0;
273: }
274: data1_start[proc_id]++;
275: }
276: }
278: for (i=0; i<is_max; i++) {
279: ISGetIndices(is[i],&idx_i);
280: for (j=0; j<n[i]; j++) {
281: idx = idx_i[j];
282: *data1_start[rank] = idx; data1_start[rank]++; /* for local proccessing */
283: proc_end = ctable[idx];
284: for (proc_id=0; proc_id<=proc_end; proc_id++) { /* for others to process */
285: if (proc_id == rank) continue; /* done before this loop */
286: if (proc_id < proc_end && !PetscBTLookup(table[proc_id],idx)) continue; /* no need for sending idx to [proc_id] */
287: *data1_start[proc_id] = idx; data1_start[proc_id]++;
288: len_s[proc_id]++;
289: }
290: }
291: /* update header data */
292: for (proc_id=0; proc_id<size; proc_id++) {
293: if (proc_id== rank) continue;
294: *(data1 + proc_id*len + 1 + i) = len_s[proc_id] - iwork[proc_id];
295: iwork[proc_id] = len_s[proc_id];
296: }
297: ISRestoreIndices(is[i],&idx_i);
298: }
300: nrqs = 0; nrqr = 0;
301: for (i=0; i<size; i++) {
302: data1_start[i] = data1 + i*len;
303: if (len_s[i]) {
304: nrqs++;
305: len_s[i] += 1 + is_max; /* add no. of header msg */
306: }
307: }
309: for (i=0; i<is_max; i++) {
310: ISDestroy(&is[i]);
311: }
312: PetscFree(n);
313: PetscFree(ctable);
315: /* Determine the number of messages to expect, their lengths, from from-ids */
316: PetscGatherNumberOfMessages(comm,NULL,len_s,&nrqr);
317: PetscGatherMessageLengths(comm,nrqs,nrqr,len_s,&id_r1,&len_r1);
319: /* Now post the sends */
320: PetscMalloc2(size,&s_waits1,size,&s_waits2);
321: k = 0;
322: for (proc_id=0; proc_id<size; proc_id++) { /* send data1 to processor [proc_id] */
323: if (len_s[proc_id]) {
324: MPI_Isend(data1_start[proc_id],len_s[proc_id],MPIU_INT,proc_id,tag1,comm,s_waits1+k);
325: k++;
326: }
327: }
329: /* 2. Receive other's is[] and process. Then send back */
330: /*-----------------------------------------------------*/
331: len = 0;
332: for (i=0; i<nrqr; i++) {
333: if (len_r1[i] > len) len = len_r1[i];
334: }
335: PetscFree(len_r1);
336: PetscFree(id_r1);
338: for (proc_id=0; proc_id<size; proc_id++) len_s[proc_id] = iwork[proc_id] = 0;
340: PetscMalloc1(len+1,&odata1);
341: PetscMalloc1(size,&odata2_ptr);
342: PetscBTCreate(Mbs,&otable);
344: len_max = ois_max*(Mbs+1); /* max space storing all is[] for each receive */
345: len_est = 2*len_max; /* estimated space of storing is[] for all receiving messages */
346: PetscMalloc1(len_est+1,&odata2);
347: nodata2 = 0; /* nodata2+1: num of PetscMalloc(,&odata2_ptr[]) called */
349: odata2_ptr[nodata2] = odata2;
351: len_unused = len_est; /* unused space in the array odata2_ptr[nodata2]-- needs to be >= len_max */
353: k = 0;
354: while (k < nrqr) {
355: /* Receive messages */
356: MPI_Iprobe(MPI_ANY_SOURCE,tag1,comm,&flag,&r_status);
357: if (flag) {
358: MPI_Get_count(&r_status,MPIU_INT,&len);
359: proc_id = r_status.MPI_SOURCE;
360: MPI_Irecv(odata1,len,MPIU_INT,proc_id,r_status.MPI_TAG,comm,&r_req);
361: MPI_Wait(&r_req,&r_status);
363: /* Process messages */
364: /* make sure there is enough unused space in odata2 array */
365: if (len_unused < len_max) { /* allocate more space for odata2 */
366: PetscMalloc1(len_est+1,&odata2);
368: odata2_ptr[++nodata2] = odata2;
370: len_unused = len_est;
371: }
373: MatIncreaseOverlap_MPISBAIJ_Local(C,odata1,OTHER,odata2,&otable);
374: len = 1 + odata2[0];
375: for (i=0; i<odata2[0]; i++) len += odata2[1 + i];
377: /* Send messages back */
378: MPI_Isend(odata2,len,MPIU_INT,proc_id,tag2,comm,s_waits2+k);
379: k++;
380: odata2 += len;
381: len_unused -= len;
382: len_s[proc_id] = len; /* num of messages sending back to [proc_id] by this proc */
383: }
384: }
385: PetscFree(odata1);
386: PetscBTDestroy(&otable);
388: /* 3. Do local work on this processor's is[] */
389: /*-------------------------------------------*/
390: /* make sure there is enough unused space in odata2(=data) array */
391: len_max = is_max*(Mbs+1); /* max space storing all is[] for this processor */
392: if (len_unused < len_max) { /* allocate more space for odata2 */
393: PetscMalloc1(len_est+1,&odata2);
395: odata2_ptr[++nodata2] = odata2;
396: }
398: data = odata2;
399: MatIncreaseOverlap_MPISBAIJ_Local(C,data1_start[rank],MINE,data,table);
400: PetscFree(data1_start);
402: /* 4. Receive work done on other processors, then merge */
403: /*------------------------------------------------------*/
404: /* get max number of messages that this processor expects to recv */
405: MPI_Allreduce(len_s,iwork,size,MPI_INT,MPI_MAX,comm);
406: PetscMalloc1(iwork[rank]+1,&data2);
407: PetscFree4(len_s,btable,iwork,Bowners);
409: k = 0;
410: while (k < nrqs) {
411: /* Receive messages */
412: MPI_Iprobe(MPI_ANY_SOURCE,tag2,comm,&flag,&r_status);
413: if (flag) {
414: MPI_Get_count(&r_status,MPIU_INT,&len);
416: proc_id = r_status.MPI_SOURCE;
418: MPI_Irecv(data2,len,MPIU_INT,proc_id,r_status.MPI_TAG,comm,&r_req);
419: MPI_Wait(&r_req,&r_status);
420: if (len > 1+is_max) { /* Add data2 into data */
421: data2_i = data2 + 1 + is_max;
422: for (i=0; i<is_max; i++) {
423: table_i = table[i];
424: data_i = data + 1 + is_max + Mbs*i;
425: isz = data[1+i];
426: for (j=0; j<data2[1+i]; j++) {
427: col = data2_i[j];
428: if (!PetscBTLookupSet(table_i,col)) data_i[isz++] = col;
429: }
430: data[1+i] = isz;
431: if (i < is_max - 1) data2_i += data2[1+i];
432: }
433: }
434: k++;
435: }
436: }
437: PetscFree(data2);
438: PetscFree2(table,t_p);
440: /* phase 1 sends are complete */
441: PetscMalloc1(size,&s_status);
442: if (nrqs) {MPI_Waitall(nrqs,s_waits1,s_status);}
443: PetscFree(data1);
445: /* phase 2 sends are complete */
446: if (nrqr) {MPI_Waitall(nrqr,s_waits2,s_status);}
447: PetscFree2(s_waits1,s_waits2);
448: PetscFree(s_status);
450: /* 5. Create new is[] */
451: /*--------------------*/
452: for (i=0; i<is_max; i++) {
453: data_i = data + 1 + is_max + Mbs*i;
454: ISCreateGeneral(PETSC_COMM_SELF,data[1+i],data_i,PETSC_COPY_VALUES,is+i);
455: }
456: for (k=0; k<=nodata2; k++) {
457: PetscFree(odata2_ptr[k]);
458: }
459: PetscFree(odata2_ptr);
460: return(0);
461: }
465: /*
466: MatIncreaseOverlap_MPISBAIJ_Local - Called by MatIncreaseOverlap, to do
467: the work on the local processor.
469: Inputs:
470: C - MAT_MPISBAIJ;
471: data - holds is[]. See MatIncreaseOverlap_MPISBAIJ_Once() for the format.
472: whose - whose is[] to be processed,
473: MINE: this processor's is[]
474: OTHER: other processor's is[]
475: Output:
476: nidx - whose = MINE:
477: holds input and newly found indices in the same format as data
478: whose = OTHER:
479: only holds the newly found indices
480: table - table[i]: mark the indices of is[i], i=0,...,is_max. Used only in the case 'whose=MINE'.
481: */
482: /* Would computation be reduced by swapping the loop 'for each is' and 'for each row'? */
483: static PetscErrorCode MatIncreaseOverlap_MPISBAIJ_Local(Mat C,PetscInt *data,PetscInt whose,PetscInt *nidx,PetscBT *table)
484: {
485: Mat_MPISBAIJ *c = (Mat_MPISBAIJ*)C->data;
486: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)(c->A)->data;
487: Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(c->B)->data;
489: PetscInt row,mbs,Mbs,*nidx_i,col,col_max,isz,isz0,*ai,*aj,*bi,*bj,*garray,rstart,l;
490: PetscInt a_start,a_end,b_start,b_end,i,j,k,is_max,*idx_i,n;
491: PetscBT table0; /* mark the indices of input is[] for look up */
492: PetscBT table_i; /* poits to i-th table. When whose=OTHER, a single table is used for all is[] */
495: Mbs = c->Mbs; mbs = a->mbs;
496: ai = a->i; aj = a->j;
497: bi = b->i; bj = b->j;
498: garray = c->garray;
499: rstart = c->rstartbs;
500: is_max = data[0];
502: PetscBTCreate(Mbs,&table0);
504: nidx[0] = is_max;
505: idx_i = data + is_max + 1; /* ptr to input is[0] array */
506: nidx_i = nidx + is_max + 1; /* ptr to output is[0] array */
507: for (i=0; i<is_max; i++) { /* for each is */
508: isz = 0;
509: n = data[1+i]; /* size of input is[i] */
511: /* initialize and set table_i(mark idx and nidx) and table0(only mark idx) */
512: if (whose == MINE) { /* process this processor's is[] */
513: table_i = table[i];
514: nidx_i = nidx + 1+ is_max + Mbs*i;
515: } else { /* process other processor's is[] - only use one temp table */
516: table_i = table[0];
517: }
518: PetscBTMemzero(Mbs,table_i);
519: PetscBTMemzero(Mbs,table0);
520: if (n==0) {
521: nidx[1+i] = 0; /* size of new is[i] */
522: continue;
523: }
525: isz0 = 0; col_max = 0;
526: for (j=0; j<n; j++) {
527: col = idx_i[j];
528: if (col >= Mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"index col %D >= Mbs %D",col,Mbs);
529: if (!PetscBTLookupSet(table_i,col)) {
530: PetscBTSet(table0,col);
531: if (whose == MINE) nidx_i[isz0] = col;
532: if (col_max < col) col_max = col;
533: isz0++;
534: }
535: }
537: if (whose == MINE) isz = isz0;
538: k = 0; /* no. of indices from input is[i] that have been examined */
539: for (row=0; row<mbs; row++) {
540: a_start = ai[row]; a_end = ai[row+1];
541: b_start = bi[row]; b_end = bi[row+1];
542: if (PetscBTLookup(table0,row+rstart)) { /* row is on input is[i]:
543: do row search: collect all col in this row */
544: for (l = a_start; l<a_end ; l++) { /* Amat */
545: col = aj[l] + rstart;
546: if (!PetscBTLookupSet(table_i,col)) nidx_i[isz++] = col;
547: }
548: for (l = b_start; l<b_end ; l++) { /* Bmat */
549: col = garray[bj[l]];
550: if (!PetscBTLookupSet(table_i,col)) nidx_i[isz++] = col;
551: }
552: k++;
553: if (k >= isz0) break; /* for (row=0; row<mbs; row++) */
554: } else { /* row is not on input is[i]:
555: do col serach: add row onto nidx_i if there is a col in nidx_i */
556: for (l = a_start; l<a_end; l++) { /* Amat */
557: col = aj[l] + rstart;
558: if (col > col_max) break;
559: if (PetscBTLookup(table0,col)) {
560: if (!PetscBTLookupSet(table_i,row+rstart)) nidx_i[isz++] = row+rstart;
561: break; /* for l = start; l<end ; l++) */
562: }
563: }
564: for (l = b_start; l<b_end; l++) { /* Bmat */
565: col = garray[bj[l]];
566: if (col > col_max) break;
567: if (PetscBTLookup(table0,col)) {
568: if (!PetscBTLookupSet(table_i,row+rstart)) nidx_i[isz++] = row+rstart;
569: break; /* for l = start; l<end ; l++) */
570: }
571: }
572: }
573: }
575: if (i < is_max - 1) {
576: idx_i += n; /* ptr to input is[i+1] array */
577: nidx_i += isz; /* ptr to output is[i+1] array */
578: }
579: nidx[1+i] = isz; /* size of new is[i] */
580: } /* for each is */
581: PetscBTDestroy(&table0);
582: return(0);
583: }