Actual source code: ispai.c
petsc-3.9.4 2018-09-11
2: /*
3: 3/99 Modified by Stephen Barnard to support SPAI version 3.0
4: */
6: /*
7: Provides an interface to the SPAI Sparse Approximate Inverse Preconditioner
8: Code written by Stephen Barnard.
10: Note: there is some BAD memory bleeding below!
12: This code needs work
14: 1) get rid of all memory bleeding
15: 2) fix PETSc/interface so that it gets if the matrix is symmetric from the matrix
16: rather than having the sp flag for PC_SPAI
17: 3) fix to set the block size based on the matrix block size
19: */
20: #define PETSC_SKIP_COMPLEX /* since spai uses I which conflicts with some complex implementations */
22: #include <petsc/private/pcimpl.h>
23: #include <../src/ksp/pc/impls/spai/petscspai.h>
25: /*
26: These are the SPAI include files
27: */
28: EXTERN_C_BEGIN
29: #define SPAI_USE_MPI /* required for setting SPAI_Comm correctly in basics.h */
30: #include <spai.h>
31: #include <matrix.h>
32: EXTERN_C_END
34: extern PetscErrorCode ConvertMatToMatrix(MPI_Comm,Mat,Mat,matrix**);
35: extern PetscErrorCode ConvertMatrixToMat(MPI_Comm,matrix*,Mat*);
36: extern PetscErrorCode ConvertVectorToVec(MPI_Comm,vector *v,Vec *Pv);
37: extern PetscErrorCode MM_to_PETSC(char*,char*,char*);
39: typedef struct {
41: matrix *B; /* matrix in SPAI format */
42: matrix *BT; /* transpose of matrix in SPAI format */
43: matrix *M; /* the approximate inverse in SPAI format */
45: Mat PM; /* the approximate inverse PETSc format */
47: double epsilon; /* tolerance */
48: int nbsteps; /* max number of "improvement" steps per line */
49: int max; /* max dimensions of is_I, q, etc. */
50: int maxnew; /* max number of new entries per step */
51: int block_size; /* constant block size */
52: int cache_size; /* one of (1,2,3,4,5,6) indicting size of cache */
53: int verbose; /* SPAI prints timing and statistics */
55: int sp; /* symmetric nonzero pattern */
56: MPI_Comm comm_spai; /* communicator to be used with spai */
57: } PC_SPAI;
59: /**********************************************************************/
61: static PetscErrorCode PCSetUp_SPAI(PC pc)
62: {
63: PC_SPAI *ispai = (PC_SPAI*)pc->data;
65: Mat AT;
68: init_SPAI();
70: if (ispai->sp) {
71: ConvertMatToMatrix(ispai->comm_spai,pc->pmat,pc->pmat,&ispai->B);
72: } else {
73: /* Use the transpose to get the column nonzero structure. */
74: MatTranspose(pc->pmat,MAT_INITIAL_MATRIX,&AT);
75: ConvertMatToMatrix(ispai->comm_spai,pc->pmat,AT,&ispai->B);
76: MatDestroy(&AT);
77: }
79: /* Destroy the transpose */
80: /* Don't know how to do it. PETSc developers? */
82: /* construct SPAI preconditioner */
83: /* FILE *messages */ /* file for warning messages */
84: /* double epsilon */ /* tolerance */
85: /* int nbsteps */ /* max number of "improvement" steps per line */
86: /* int max */ /* max dimensions of is_I, q, etc. */
87: /* int maxnew */ /* max number of new entries per step */
88: /* int block_size */ /* block_size == 1 specifies scalar elments
89: block_size == n specifies nxn constant-block elements
90: block_size == 0 specifies variable-block elements */
91: /* int cache_size */ /* one of (1,2,3,4,5,6) indicting size of cache. cache_size == 0 indicates no caching */
92: /* int verbose */ /* verbose == 0 specifies that SPAI is silent
93: verbose == 1 prints timing and matrix statistics */
95: bspai(ispai->B,&ispai->M,
96: stdout,
97: ispai->epsilon,
98: ispai->nbsteps,
99: ispai->max,
100: ispai->maxnew,
101: ispai->block_size,
102: ispai->cache_size,
103: ispai->verbose);
105: ConvertMatrixToMat(PetscObjectComm((PetscObject)pc),ispai->M,&ispai->PM);
107: /* free the SPAI matrices */
108: sp_free_matrix(ispai->B);
109: sp_free_matrix(ispai->M);
110: return(0);
111: }
113: /**********************************************************************/
115: static PetscErrorCode PCApply_SPAI(PC pc,Vec xx,Vec y)
116: {
117: PC_SPAI *ispai = (PC_SPAI*)pc->data;
121: /* Now using PETSc's multiply */
122: MatMult(ispai->PM,xx,y);
123: return(0);
124: }
126: /**********************************************************************/
128: static PetscErrorCode PCDestroy_SPAI(PC pc)
129: {
131: PC_SPAI *ispai = (PC_SPAI*)pc->data;
134: MatDestroy(&ispai->PM);
135: MPI_Comm_free(&(ispai->comm_spai));
136: PetscFree(pc->data);
137: return(0);
138: }
140: /**********************************************************************/
142: static PetscErrorCode PCView_SPAI(PC pc,PetscViewer viewer)
143: {
144: PC_SPAI *ispai = (PC_SPAI*)pc->data;
146: PetscBool iascii;
149: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
150: if (iascii) {
151: PetscViewerASCIIPrintf(viewer," epsilon %g\n", (double)ispai->epsilon);
152: PetscViewerASCIIPrintf(viewer," nbsteps %d\n", ispai->nbsteps);
153: PetscViewerASCIIPrintf(viewer," max %d\n", ispai->max);
154: PetscViewerASCIIPrintf(viewer," maxnew %d\n", ispai->maxnew);
155: PetscViewerASCIIPrintf(viewer," block_size %d\n",ispai->block_size);
156: PetscViewerASCIIPrintf(viewer," cache_size %d\n",ispai->cache_size);
157: PetscViewerASCIIPrintf(viewer," verbose %d\n", ispai->verbose);
158: PetscViewerASCIIPrintf(viewer," sp %d\n", ispai->sp);
159: }
160: return(0);
161: }
163: static PetscErrorCode PCSPAISetEpsilon_SPAI(PC pc,double epsilon1)
164: {
165: PC_SPAI *ispai = (PC_SPAI*)pc->data;
168: ispai->epsilon = epsilon1;
169: return(0);
170: }
172: /**********************************************************************/
174: static PetscErrorCode PCSPAISetNBSteps_SPAI(PC pc,int nbsteps1)
175: {
176: PC_SPAI *ispai = (PC_SPAI*)pc->data;
179: ispai->nbsteps = nbsteps1;
180: return(0);
181: }
183: /**********************************************************************/
185: /* added 1/7/99 g.h. */
186: static PetscErrorCode PCSPAISetMax_SPAI(PC pc,int max1)
187: {
188: PC_SPAI *ispai = (PC_SPAI*)pc->data;
191: ispai->max = max1;
192: return(0);
193: }
195: /**********************************************************************/
197: static PetscErrorCode PCSPAISetMaxNew_SPAI(PC pc,int maxnew1)
198: {
199: PC_SPAI *ispai = (PC_SPAI*)pc->data;
202: ispai->maxnew = maxnew1;
203: return(0);
204: }
206: /**********************************************************************/
208: static PetscErrorCode PCSPAISetBlockSize_SPAI(PC pc,int block_size1)
209: {
210: PC_SPAI *ispai = (PC_SPAI*)pc->data;
213: ispai->block_size = block_size1;
214: return(0);
215: }
217: /**********************************************************************/
219: static PetscErrorCode PCSPAISetCacheSize_SPAI(PC pc,int cache_size)
220: {
221: PC_SPAI *ispai = (PC_SPAI*)pc->data;
224: ispai->cache_size = cache_size;
225: return(0);
226: }
228: /**********************************************************************/
230: static PetscErrorCode PCSPAISetVerbose_SPAI(PC pc,int verbose)
231: {
232: PC_SPAI *ispai = (PC_SPAI*)pc->data;
235: ispai->verbose = verbose;
236: return(0);
237: }
239: /**********************************************************************/
241: static PetscErrorCode PCSPAISetSp_SPAI(PC pc,int sp)
242: {
243: PC_SPAI *ispai = (PC_SPAI*)pc->data;
246: ispai->sp = sp;
247: return(0);
248: }
250: /* -------------------------------------------------------------------*/
252: /*@
253: PCSPAISetEpsilon -- Set the tolerance for the SPAI preconditioner
255: Input Parameters:
256: + pc - the preconditioner
257: - eps - epsilon (default .4)
259: Notes: Espilon must be between 0 and 1. It controls the
260: quality of the approximation of M to the inverse of
261: A. Higher values of epsilon lead to more work, more
262: fill, and usually better preconditioners. In many
263: cases the best choice of epsilon is the one that
264: divides the total solution time equally between the
265: preconditioner and the solver.
267: Level: intermediate
269: .seealso: PCSPAI, PCSetType()
270: @*/
271: PetscErrorCode PCSPAISetEpsilon(PC pc,double epsilon1)
272: {
276: PetscTryMethod(pc,"PCSPAISetEpsilon_C",(PC,double),(pc,epsilon1));
277: return(0);
278: }
280: /**********************************************************************/
282: /*@
283: PCSPAISetNBSteps - set maximum number of improvement steps per row in
284: the SPAI preconditioner
286: Input Parameters:
287: + pc - the preconditioner
288: - n - number of steps (default 5)
290: Notes: SPAI constructs to approximation to every column of
291: the exact inverse of A in a series of improvement
292: steps. The quality of the approximation is determined
293: by epsilon. If an approximation achieving an accuracy
294: of epsilon is not obtained after ns steps, SPAI simply
295: uses the best approximation constructed so far.
297: Level: intermediate
299: .seealso: PCSPAI, PCSetType(), PCSPAISetMaxNew()
300: @*/
301: PetscErrorCode PCSPAISetNBSteps(PC pc,int nbsteps1)
302: {
306: PetscTryMethod(pc,"PCSPAISetNBSteps_C",(PC,int),(pc,nbsteps1));
307: return(0);
308: }
310: /**********************************************************************/
312: /* added 1/7/99 g.h. */
313: /*@
314: PCSPAISetMax - set the size of various working buffers in
315: the SPAI preconditioner
317: Input Parameters:
318: + pc - the preconditioner
319: - n - size (default is 5000)
321: Level: intermediate
323: .seealso: PCSPAI, PCSetType()
324: @*/
325: PetscErrorCode PCSPAISetMax(PC pc,int max1)
326: {
330: PetscTryMethod(pc,"PCSPAISetMax_C",(PC,int),(pc,max1));
331: return(0);
332: }
334: /**********************************************************************/
336: /*@
337: PCSPAISetMaxNew - set maximum number of new nonzero candidates per step
338: in SPAI preconditioner
340: Input Parameters:
341: + pc - the preconditioner
342: - n - maximum number (default 5)
344: Level: intermediate
346: .seealso: PCSPAI, PCSetType(), PCSPAISetNBSteps()
347: @*/
348: PetscErrorCode PCSPAISetMaxNew(PC pc,int maxnew1)
349: {
353: PetscTryMethod(pc,"PCSPAISetMaxNew_C",(PC,int),(pc,maxnew1));
354: return(0);
355: }
357: /**********************************************************************/
359: /*@
360: PCSPAISetBlockSize - set the block size for the SPAI preconditioner
362: Input Parameters:
363: + pc - the preconditioner
364: - n - block size (default 1)
366: Notes: A block
367: size of 1 treats A as a matrix of scalar elements. A
368: block size of s > 1 treats A as a matrix of sxs
369: blocks. A block size of 0 treats A as a matrix with
370: variable sized blocks, which are determined by
371: searching for dense square diagonal blocks in A.
372: This can be very effective for finite-element
373: matrices.
375: SPAI will convert A to block form, use a block
376: version of the preconditioner algorithm, and then
377: convert the result back to scalar form.
379: In many cases the a block-size parameter other than 1
380: can lead to very significant improvement in
381: performance.
384: Level: intermediate
386: .seealso: PCSPAI, PCSetType()
387: @*/
388: PetscErrorCode PCSPAISetBlockSize(PC pc,int block_size1)
389: {
393: PetscTryMethod(pc,"PCSPAISetBlockSize_C",(PC,int),(pc,block_size1));
394: return(0);
395: }
397: /**********************************************************************/
399: /*@
400: PCSPAISetCacheSize - specify cache size in the SPAI preconditioner
402: Input Parameters:
403: + pc - the preconditioner
404: - n - cache size {0,1,2,3,4,5} (default 5)
406: Notes: SPAI uses a hash table to cache messages and avoid
407: redundant communication. If suggest always using
408: 5. This parameter is irrelevant in the serial
409: version.
411: Level: intermediate
413: .seealso: PCSPAI, PCSetType()
414: @*/
415: PetscErrorCode PCSPAISetCacheSize(PC pc,int cache_size)
416: {
420: PetscTryMethod(pc,"PCSPAISetCacheSize_C",(PC,int),(pc,cache_size));
421: return(0);
422: }
424: /**********************************************************************/
426: /*@
427: PCSPAISetVerbose - verbosity level for the SPAI preconditioner
429: Input Parameters:
430: + pc - the preconditioner
431: - n - level (default 1)
433: Notes: print parameters, timings and matrix statistics
435: Level: intermediate
437: .seealso: PCSPAI, PCSetType()
438: @*/
439: PetscErrorCode PCSPAISetVerbose(PC pc,int verbose)
440: {
444: PetscTryMethod(pc,"PCSPAISetVerbose_C",(PC,int),(pc,verbose));
445: return(0);
446: }
448: /**********************************************************************/
450: /*@
451: PCSPAISetSp - specify a symmetric matrix sparsity pattern in the SPAI preconditioner
453: Input Parameters:
454: + pc - the preconditioner
455: - n - 0 or 1
457: Notes: If A has a symmetric nonzero pattern use -sp 1 to
458: improve performance by eliminating some communication
459: in the parallel version. Even if A does not have a
460: symmetric nonzero pattern -sp 1 may well lead to good
461: results, but the code will not follow the published
462: SPAI algorithm exactly.
465: Level: intermediate
467: .seealso: PCSPAI, PCSetType()
468: @*/
469: PetscErrorCode PCSPAISetSp(PC pc,int sp)
470: {
474: PetscTryMethod(pc,"PCSPAISetSp_C",(PC,int),(pc,sp));
475: return(0);
476: }
478: /**********************************************************************/
480: /**********************************************************************/
482: static PetscErrorCode PCSetFromOptions_SPAI(PetscOptionItems *PetscOptionsObject,PC pc)
483: {
484: PC_SPAI *ispai = (PC_SPAI*)pc->data;
486: int nbsteps1,max1,maxnew1,block_size1,cache_size,verbose,sp;
487: double epsilon1;
488: PetscBool flg;
491: PetscOptionsHead(PetscOptionsObject,"SPAI options");
492: PetscOptionsReal("-pc_spai_epsilon","","PCSPAISetEpsilon",ispai->epsilon,&epsilon1,&flg);
493: if (flg) {
494: PCSPAISetEpsilon(pc,epsilon1);
495: }
496: PetscOptionsInt("-pc_spai_nbsteps","","PCSPAISetNBSteps",ispai->nbsteps,&nbsteps1,&flg);
497: if (flg) {
498: PCSPAISetNBSteps(pc,nbsteps1);
499: }
500: /* added 1/7/99 g.h. */
501: PetscOptionsInt("-pc_spai_max","","PCSPAISetMax",ispai->max,&max1,&flg);
502: if (flg) {
503: PCSPAISetMax(pc,max1);
504: }
505: PetscOptionsInt("-pc_spai_maxnew","","PCSPAISetMaxNew",ispai->maxnew,&maxnew1,&flg);
506: if (flg) {
507: PCSPAISetMaxNew(pc,maxnew1);
508: }
509: PetscOptionsInt("-pc_spai_block_size","","PCSPAISetBlockSize",ispai->block_size,&block_size1,&flg);
510: if (flg) {
511: PCSPAISetBlockSize(pc,block_size1);
512: }
513: PetscOptionsInt("-pc_spai_cache_size","","PCSPAISetCacheSize",ispai->cache_size,&cache_size,&flg);
514: if (flg) {
515: PCSPAISetCacheSize(pc,cache_size);
516: }
517: PetscOptionsInt("-pc_spai_verbose","","PCSPAISetVerbose",ispai->verbose,&verbose,&flg);
518: if (flg) {
519: PCSPAISetVerbose(pc,verbose);
520: }
521: PetscOptionsInt("-pc_spai_sp","","PCSPAISetSp",ispai->sp,&sp,&flg);
522: if (flg) {
523: PCSPAISetSp(pc,sp);
524: }
525: PetscOptionsTail();
526: return(0);
527: }
529: /**********************************************************************/
531: /*MC
532: PCSPAI - Use the Sparse Approximate Inverse method of Grote and Barnard
533: as a preconditioner (SIAM J. Sci. Comput.; vol 18, nr 3)
535: Options Database Keys:
536: + -pc_spai_epsilon <eps> - set tolerance
537: . -pc_spai_nbstep <n> - set nbsteps
538: . -pc_spai_max <m> - set max
539: . -pc_spai_max_new <m> - set maxnew
540: . -pc_spai_block_size <n> - set block size
541: . -pc_spai_cache_size <n> - set cache size
542: . -pc_spai_sp <m> - set sp
543: - -pc_spai_set_verbose <true,false> - verbose output
545: Notes: This only works with AIJ matrices.
547: Level: beginner
549: Concepts: approximate inverse
551: .seealso: PCCreate(), PCSetType(), PCType (for list of available types), PC,
552: PCSPAISetEpsilon(), PCSPAISetMax(), PCSPAISetMaxNew(), PCSPAISetBlockSize(),
553: PCSPAISetVerbose(), PCSPAISetSp()
554: M*/
556: PETSC_EXTERN PetscErrorCode PCCreate_SPAI(PC pc)
557: {
558: PC_SPAI *ispai;
562: PetscNewLog(pc,&ispai);
563: pc->data = ispai;
565: pc->ops->destroy = PCDestroy_SPAI;
566: pc->ops->apply = PCApply_SPAI;
567: pc->ops->applyrichardson = 0;
568: pc->ops->setup = PCSetUp_SPAI;
569: pc->ops->view = PCView_SPAI;
570: pc->ops->setfromoptions = PCSetFromOptions_SPAI;
572: ispai->epsilon = .4;
573: ispai->nbsteps = 5;
574: ispai->max = 5000;
575: ispai->maxnew = 5;
576: ispai->block_size = 1;
577: ispai->cache_size = 5;
578: ispai->verbose = 0;
580: ispai->sp = 1;
581: MPI_Comm_dup(PetscObjectComm((PetscObject)pc),&(ispai->comm_spai));
583: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetEpsilon_C",PCSPAISetEpsilon_SPAI);
584: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetNBSteps_C",PCSPAISetNBSteps_SPAI);
585: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetMax_C",PCSPAISetMax_SPAI);
586: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetMaxNew_C",PCSPAISetMaxNew_SPAI);
587: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetBlockSize_C",PCSPAISetBlockSize_SPAI);
588: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetCacheSize_C",PCSPAISetCacheSize_SPAI);
589: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetVerbose_C",PCSPAISetVerbose_SPAI);
590: PetscObjectComposeFunction((PetscObject)pc,"PCSPAISetSp_C",PCSPAISetSp_SPAI);
591: return(0);
592: }
594: /**********************************************************************/
596: /*
597: Converts from a PETSc matrix to an SPAI matrix
598: */
599: PetscErrorCode ConvertMatToMatrix(MPI_Comm comm, Mat A,Mat AT,matrix **B)
600: {
601: matrix *M;
602: int i,j,col;
603: int row_indx;
604: int len,pe,local_indx,start_indx;
605: int *mapping;
606: PetscErrorCode ierr;
607: const int *cols;
608: const double *vals;
609: int n,mnl,nnl,nz,rstart,rend;
610: PetscMPIInt size,rank;
611: struct compressed_lines *rows;
614: MPI_Comm_size(comm,&size);
615: MPI_Comm_rank(comm,&rank);
616: MatGetSize(A,&n,&n);
617: MatGetLocalSize(A,&mnl,&nnl);
619: /*
620: not sure why a barrier is required. commenting out
621: MPI_Barrier(comm);
622: */
624: M = new_matrix((SPAI_Comm)comm);
626: M->n = n;
627: M->bs = 1;
628: M->max_block_size = 1;
630: M->mnls = (int*)malloc(sizeof(int)*size);
631: M->start_indices = (int*)malloc(sizeof(int)*size);
632: M->pe = (int*)malloc(sizeof(int)*n);
633: M->block_sizes = (int*)malloc(sizeof(int)*n);
634: for (i=0; i<n; i++) M->block_sizes[i] = 1;
636: MPI_Allgather(&mnl,1,MPI_INT,M->mnls,1,MPI_INT,comm);
638: M->start_indices[0] = 0;
639: for (i=1; i<size; i++) M->start_indices[i] = M->start_indices[i-1] + M->mnls[i-1];
641: M->mnl = M->mnls[M->myid];
642: M->my_start_index = M->start_indices[M->myid];
644: for (i=0; i<size; i++) {
645: start_indx = M->start_indices[i];
646: for (j=0; j<M->mnls[i]; j++) M->pe[start_indx+j] = i;
647: }
649: if (AT) {
650: M->lines = new_compressed_lines(M->mnls[rank],1);
651: } else {
652: M->lines = new_compressed_lines(M->mnls[rank],0);
653: }
655: rows = M->lines;
657: /* Determine the mapping from global indices to pointers */
658: PetscMalloc1(M->n,&mapping);
659: pe = 0;
660: local_indx = 0;
661: for (i=0; i<M->n; i++) {
662: if (local_indx >= M->mnls[pe]) {
663: pe++;
664: local_indx = 0;
665: }
666: mapping[i] = local_indx + M->start_indices[pe];
667: local_indx++;
668: }
670: /*********************************************************/
671: /************** Set up the row structure *****************/
672: /*********************************************************/
674: MatGetOwnershipRange(A,&rstart,&rend);
675: for (i=rstart; i<rend; i++) {
676: row_indx = i - rstart;
677: MatGetRow(A,i,&nz,&cols,&vals);
678: /* allocate buffers */
679: rows->ptrs[row_indx] = (int*)malloc(nz*sizeof(int));
680: rows->A[row_indx] = (double*)malloc(nz*sizeof(double));
681: /* copy the matrix */
682: for (j=0; j<nz; j++) {
683: col = cols[j];
684: len = rows->len[row_indx]++;
686: rows->ptrs[row_indx][len] = mapping[col];
687: rows->A[row_indx][len] = vals[j];
688: }
689: rows->slen[row_indx] = rows->len[row_indx];
691: MatRestoreRow(A,i,&nz,&cols,&vals);
692: }
695: /************************************************************/
696: /************** Set up the column structure *****************/
697: /*********************************************************/
699: if (AT) {
701: for (i=rstart; i<rend; i++) {
702: row_indx = i - rstart;
703: MatGetRow(AT,i,&nz,&cols,&vals);
704: /* allocate buffers */
705: rows->rptrs[row_indx] = (int*)malloc(nz*sizeof(int));
706: /* copy the matrix (i.e., the structure) */
707: for (j=0; j<nz; j++) {
708: col = cols[j];
709: len = rows->rlen[row_indx]++;
711: rows->rptrs[row_indx][len] = mapping[col];
712: }
713: MatRestoreRow(AT,i,&nz,&cols,&vals);
714: }
715: }
717: PetscFree(mapping);
719: order_pointers(M);
720: M->maxnz = calc_maxnz(M);
721: *B = M;
722: return(0);
723: }
725: /**********************************************************************/
727: /*
728: Converts from an SPAI matrix B to a PETSc matrix PB.
729: This assumes that the SPAI matrix B is stored in
730: COMPRESSED-ROW format.
731: */
732: PetscErrorCode ConvertMatrixToMat(MPI_Comm comm,matrix *B,Mat *PB)
733: {
734: PetscMPIInt size,rank;
736: int m,n,M,N;
737: int d_nz,o_nz;
738: int *d_nnz,*o_nnz;
739: int i,k,global_row,global_col,first_diag_col,last_diag_col;
740: PetscScalar val;
743: MPI_Comm_size(comm,&size);
744: MPI_Comm_rank(comm,&rank);
746: m = n = B->mnls[rank];
747: d_nz = o_nz = 0;
749: /* Determine preallocation for MatCreateMPIAIJ */
750: PetscMalloc1(m,&d_nnz);
751: PetscMalloc1(m,&o_nnz);
752: for (i=0; i<m; i++) d_nnz[i] = o_nnz[i] = 0;
753: first_diag_col = B->start_indices[rank];
754: last_diag_col = first_diag_col + B->mnls[rank];
755: for (i=0; i<B->mnls[rank]; i++) {
756: for (k=0; k<B->lines->len[i]; k++) {
757: global_col = B->lines->ptrs[i][k];
758: if ((global_col >= first_diag_col) && (global_col < last_diag_col)) d_nnz[i]++;
759: else o_nnz[i]++;
760: }
761: }
763: M = N = B->n;
764: /* Here we only know how to create AIJ format */
765: MatCreate(comm,PB);
766: MatSetSizes(*PB,m,n,M,N);
767: MatSetType(*PB,MATAIJ);
768: MatSeqAIJSetPreallocation(*PB,d_nz,d_nnz);
769: MatMPIAIJSetPreallocation(*PB,d_nz,d_nnz,o_nz,o_nnz);
771: for (i=0; i<B->mnls[rank]; i++) {
772: global_row = B->start_indices[rank]+i;
773: for (k=0; k<B->lines->len[i]; k++) {
774: global_col = B->lines->ptrs[i][k];
776: val = B->lines->A[i][k];
777: MatSetValues(*PB,1,&global_row,1,&global_col,&val,ADD_VALUES);
778: }
779: }
781: PetscFree(d_nnz);
782: PetscFree(o_nnz);
784: MatAssemblyBegin(*PB,MAT_FINAL_ASSEMBLY);
785: MatAssemblyEnd(*PB,MAT_FINAL_ASSEMBLY);
786: return(0);
787: }
789: /**********************************************************************/
791: /*
792: Converts from an SPAI vector v to a PETSc vec Pv.
793: */
794: PetscErrorCode ConvertVectorToVec(MPI_Comm comm,vector *v,Vec *Pv)
795: {
797: PetscMPIInt size,rank;
798: int m,M,i,*mnls,*start_indices,*global_indices;
801: MPI_Comm_size(comm,&size);
802: MPI_Comm_rank(comm,&rank);
804: m = v->mnl;
805: M = v->n;
808: VecCreateMPI(comm,m,M,Pv);
810: PetscMalloc1(size,&mnls);
811: MPI_Allgather(&v->mnl,1,MPI_INT,mnls,1,MPI_INT,comm);
813: PetscMalloc1(size,&start_indices);
815: start_indices[0] = 0;
816: for (i=1; i<size; i++) start_indices[i] = start_indices[i-1] +mnls[i-1];
818: PetscMalloc1(v->mnl,&global_indices);
819: for (i=0; i<v->mnl; i++) global_indices[i] = start_indices[rank] + i;
821: PetscFree(mnls);
822: PetscFree(start_indices);
824: VecSetValues(*Pv,v->mnl,global_indices,v->v,INSERT_VALUES);
825: VecAssemblyBegin(*Pv);
826: VecAssemblyEnd(*Pv);
828: PetscFree(global_indices);
829: return(0);
830: }