Actual source code: party.c
petsc-3.4.5 2014-06-29
2: #include <../src/mat/impls/adj/mpi/mpiadj.h> /*I "petscmat.h" I*/
4: #if defined(PETSC_HAVE_UNISTD_H)
5: #include <unistd.h>
6: #endif
8: /*
9: Currently using Party-1.99
10: */
11: EXTERN_C_BEGIN
12: #include <party_lib.h>
13: EXTERN_C_END
15: typedef struct {
16: PetscBool redm;
17: PetscBool redo;
18: PetscBool recursive;
19: PetscBool verbose;
20: char global[15]; /* global method */
21: char local[15]; /* local method */
22: PetscInt nbvtxcoarsed; /* number of vertices for the coarse graph */
23: } MatPartitioning_Party;
25: #define SIZE_LOG 10000 /* size of buffer for mesg_log */
29: static PetscErrorCode MatPartitioningApply_Party(MatPartitioning part,IS *partitioning)
30: {
31: PetscErrorCode ierr;
32: PetscInt i,*parttab,*locals,nb_locals,M,N;
33: PetscMPIInt size,rank;
34: Mat mat = part->adj,matAdj,matSeq,*A;
35: Mat_MPIAdj *adj;
36: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
37: PetscBool flg;
38: IS isrow, iscol;
39: int n,*edge_p,*edge,*vertex_w,p,*part_party,cutsize,redl,rec;
40: const char *redm,*redo;
41: char *mesg_log;
42: #if defined(PETSC_HAVE_UNISTD_H)
43: int fd_stdout,fd_pipe[2],count,err;
44: #endif
47: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
48: MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&rank);
49: PetscObjectTypeCompare((PetscObject)mat,MATMPIADJ,&flg);
50: if (size>1) {
51: if (flg) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Distributed matrix format MPIAdj is not supported for sequential partitioners");
52: PetscInfo(part,"Converting distributed matrix to sequential: this could be a performance loss\n");
53: MatGetSize(mat,&M,&N);
54: ISCreateStride(PETSC_COMM_SELF,M,0,1,&isrow);
55: ISCreateStride(PETSC_COMM_SELF,N,0,1,&iscol);
56: MatGetSubMatrices(mat,1,&isrow,&iscol,MAT_INITIAL_MATRIX,&A);
57: ISDestroy(&isrow);
58: ISDestroy(&iscol);
59: matSeq = *A;
60: PetscFree(A);
61: } else {
62: PetscObjectReference((PetscObject)mat);
63: matSeq = mat;
64: }
66: if (!flg) { /* convert regular matrix to MPIADJ */
67: MatConvert(matSeq,MATMPIADJ,MAT_INITIAL_MATRIX,&matAdj);
68: } else {
69: PetscObjectReference((PetscObject)matSeq);
70: matAdj = matSeq;
71: }
73: adj = (Mat_MPIAdj*)matAdj->data; /* finaly adj contains adjacency graph */
75: /* arguments for Party library */
76: n = mat->rmap->N; /* number of vertices in full graph */
77: edge_p = adj->i; /* start of edge list for each vertex */
78: edge = adj->j; /* edge list data */
79: vertex_w = part->vertex_weights; /* weights for all vertices */
80: p = part->n; /* number of parts to create */
81: redl = party->nbvtxcoarsed; /* how many vertices to coarsen down to? */
82: rec = party->recursive ? 1 : 0; /* recursive bisection */
83: redm = party->redm ? "lam" : ""; /* matching method */
84: redo = party->redo ? "w3" : ""; /* matching optimization method */
86: PetscMalloc((mat->rmap->N)*sizeof(int),&part_party);
88: /* redirect output to buffer */
89: #if defined(PETSC_HAVE_UNISTD_H)
90: fd_stdout = dup(1);
91: if (pipe(fd_pipe)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SYS,"Could not open pipe");
92: close(1);
93: dup2(fd_pipe[1],1);
94: PetscMalloc(SIZE_LOG*sizeof(char),&mesg_log);
95: #endif
97: /* library call */
98: party_lib_times_start();
99: party_lib(n,vertex_w,NULL,NULL,NULL,edge_p,edge,NULL,p,part_party,&cutsize,redl,(char*)redm,(char*)redo,party->global,party->local,rec,1);
101: party_lib_times_output(1);
102: part_info(n,vertex_w,edge_p,edge,NULL,p,part_party,1);
104: #if defined(PETSC_HAVE_UNISTD_H)
105: err = fflush(stdout);
106: if (err) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SYS,"fflush() failed on stdout");
107: count = read(fd_pipe[0],mesg_log,(SIZE_LOG-1)*sizeof(char));
108: if (count<0) count = 0;
109: mesg_log[count] = 0;
110: close(1);
111: dup2(fd_stdout,1);
112: close(fd_stdout);
113: close(fd_pipe[0]);
114: close(fd_pipe[1]);
115: if (party->verbose) {
116: PetscPrintf(PetscObjectComm((PetscObject)mat),mesg_log);
117: }
118: PetscFree(mesg_log);
119: #endif
120: if (ierr) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"Party failed");
122: PetscMalloc((mat->rmap->N)*sizeof(PetscInt),&parttab);
123: for (i=0; i<mat->rmap->N; i++) parttab[i] = part_party[i];
125: /* creation of the index set */
126: nb_locals = mat->rmap->N / size;
127: locals = parttab + rank*nb_locals;
128: if (rank < mat->rmap->N % size) {
129: nb_locals++;
130: locals += rank;
131: } else locals += mat->rmap->N % size;
133: ISCreateGeneral(PetscObjectComm((PetscObject)part),nb_locals,locals,PETSC_COPY_VALUES,partitioning);
135: /* clean up */
136: PetscFree(parttab);
137: PetscFree(part_party);
138: MatDestroy(&matSeq);
139: MatDestroy(&matAdj);
140: return(0);
141: }
145: PetscErrorCode MatPartitioningView_Party(MatPartitioning part,PetscViewer viewer)
146: {
147: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
148: PetscErrorCode ierr;
149: PetscBool isascii;
152: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isascii);
153: if (isascii) {
154: PetscViewerASCIIPrintf(viewer," Global method: %s\n",party->global);
155: PetscViewerASCIIPrintf(viewer," Local method: %s\n",party->local);
156: PetscViewerASCIIPrintf(viewer," Number of vertices for the coarse graph: %d\n",party->nbvtxcoarsed);
157: if (party->redm) {
158: PetscViewerASCIIPrintf(viewer," Using matching method for graph reduction\n");
159: }
160: if (party->redo) {
161: PetscViewerASCIIPrintf(viewer," Using matching optimization\n");
162: }
163: if (party->recursive) {
164: PetscViewerASCIIPrintf(viewer," Using recursive bipartitioning\n");
165: }
166: }
167: return(0);
168: }
172: /*@C
173: MatPartitioningPartySetGlobal - Set global method for Party partitioner.
175: Collective on MatPartitioning
177: Input Parameters:
178: + part - the partitioning context
179: - method - a string representing the method
181: Options Database:
182: . -mat_partitioning_party_global <method> - the global method
184: Level: advanced
186: Notes:
187: The method may be one of MP_PARTY_OPT, MP_PARTY_LIN, MP_PARTY_SCA,
188: MP_PARTY_RAN, MP_PARTY_GBF, MP_PARTY_GCF, MP_PARTY_BUB or MP_PARTY_DEF, or
189: alternatively a string describing the method. Two or more methods can be
190: combined like "gbf,gcf". Check the Party Library Users Manual for details.
192: .seealso: MatPartitioningPartySetLocal()
193: @*/
194: PetscErrorCode MatPartitioningPartySetGlobal(MatPartitioning part,const char *global)
195: {
200: PetscTryMethod(part,"MatPartitioningPartySetGlobal_C",(MatPartitioning,const char*),(part,global));
201: return(0);
202: }
206: PetscErrorCode MatPartitioningPartySetGlobal_Party(MatPartitioning part,const char *global)
207: {
208: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
209: PetscErrorCode ierr;
212: PetscStrncpy(party->global,global,15);
213: return(0);
214: }
218: /*@C
219: MatPartitioningPartySetLocal - Set local method for Party partitioner.
221: Collective on MatPartitioning
223: Input Parameters:
224: + part - the partitioning context
225: - method - a string representing the method
227: Options Database:
228: . -mat_partitioning_party_local <method> - the local method
230: Level: advanced
232: Notes:
233: The method may be one of MP_PARTY_HELPFUL_SETS, MP_PARTY_KERNIGHAN_LIN, or
234: MP_PARTY_NONE. Check the Party Library Users Manual for details.
236: .seealso: MatPartitioningPartySetGlobal()
237: @*/
238: PetscErrorCode MatPartitioningPartySetLocal(MatPartitioning part,const char *local)
239: {
244: PetscTryMethod(part,"MatPartitioningPartySetLocal_C",(MatPartitioning,const char*),(part,local));
245: return(0);
246: }
250: PetscErrorCode MatPartitioningPartySetLocal_Party(MatPartitioning part,const char *local)
252: {
253: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
254: PetscErrorCode ierr;
257: PetscStrncpy(party->local,local,15);
258: return(0);
259: }
263: /*@
264: MatPartitioningPartySetCoarseLevel - Set the coarse level parameter for the
265: Party partitioner.
267: Collective on MatPartitioning
269: Input Parameters:
270: + part - the partitioning context
271: - level - the coarse level in range [0.0,1.0]
273: Options Database:
274: . -mat_partitioning_party_coarse <l> - Coarse level
276: Level: advanced
277: @*/
278: PetscErrorCode MatPartitioningPartySetCoarseLevel(MatPartitioning part,PetscReal level)
279: {
285: PetscTryMethod(part,"MatPartitioningPartySetCoarseLevel_C",(MatPartitioning,PetscReal),(part,level));
286: return(0);
287: }
291: PetscErrorCode MatPartitioningPartySetCoarseLevel_Party(MatPartitioning part,PetscReal level)
292: {
293: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
296: if (level<0.0 || level>1.0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Party: level of coarsening out of range [0.0-1.0]");
297: party->nbvtxcoarsed = (PetscInt)(part->adj->cmap->N * level);
298: if (party->nbvtxcoarsed < 20) party->nbvtxcoarsed = 20;
299: return(0);
300: }
304: /*@
305: MatPartitioningPartySetMatchOptimization - Activate matching optimization for
306: graph reduction.
308: Collective on MatPartitioning
310: Input Parameters:
311: + part - the partitioning context
312: - opt - boolean flag
314: Options Database:
315: . -mat_partitioning_party_match_optimization - Matching optimization on/off
317: Level: advanced
318: @*/
319: PetscErrorCode MatPartitioningPartySetMatchOptimization(MatPartitioning part,PetscBool opt)
320: {
326: PetscTryMethod(part,"MatPartitioningPartySetMatchOptimization_C",(MatPartitioning,PetscBool),(part,opt));
327: return(0);
328: }
332: PetscErrorCode MatPartitioningPartySetMatchOptimization_Party(MatPartitioning part,PetscBool opt)
333: {
334: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
337: party->redo = opt;
338: return(0);
339: }
343: /*@
344: MatPartitioningPartySetBipart - Activate or deactivate recursive bisection.
346: Collective on MatPartitioning
348: Input Parameters:
349: + part - the partitioning context
350: - bp - boolean flag
352: Options Database:
353: - -mat_partitioning_party_bipart - Bipartitioning option on/off
355: Level: advanced
356: @*/
357: PetscErrorCode MatPartitioningPartySetBipart(MatPartitioning part,PetscBool bp)
358: {
364: PetscTryMethod(part,"MatPartitioningPartySetBipart_C",(MatPartitioning,PetscBool),(part,bp));
365: return(0);
366: }
370: PetscErrorCode MatPartitioningPartySetBipart_Party(MatPartitioning part,PetscBool bp)
371: {
372: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
375: party->recursive = bp;
376: return(0);
377: }
381: PetscErrorCode MatPartitioningSetFromOptions_Party(MatPartitioning part)
382: {
383: PetscErrorCode ierr;
384: PetscBool flag;
385: char value[256];
386: PetscReal r;
387: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
390: PetscOptionsHead("Set Party partitioning options");
391: PetscOptionsString("-mat_partitioning_party_global","Global method","MatPartitioningPartySetGlobal",party->global,value,256,&flag);
392: if (flag) { MatPartitioningPartySetGlobal(part,value); }
393: PetscOptionsString("-mat_partitioning_party_local","Local method","MatPartitioningPartySetLocal",party->local,value,256,&flag);
394: if (flag) { MatPartitioningPartySetLocal(part,value); }
395: PetscOptionsReal("-mat_partitioning_party_coarse","Coarse level","MatPartitioningPartySetCoarseLevel",0.0,&r,&flag);
396: if (flag) { MatPartitioningPartySetCoarseLevel(part,r); }
397: PetscOptionsBool("-mat_partitioning_party_match_optimization","Matching optimization on/off","MatPartitioningPartySetMatchOptimization",party->redo,&party->redo,NULL);
398: PetscOptionsBool("-mat_partitioning_party_bipart","Bipartitioning on/off","MatPartitioningPartySetBipart",party->recursive,&party->recursive,NULL);
399: PetscOptionsBool("-mat_partitioning_party_verbose","Show library output","",party->verbose,&party->verbose,NULL);
400: PetscOptionsTail();
401: return(0);
402: }
406: PetscErrorCode MatPartitioningDestroy_Party(MatPartitioning part)
407: {
408: MatPartitioning_Party *party = (MatPartitioning_Party*)part->data;
409: PetscErrorCode ierr;
412: PetscFree(party);
413: /* clear composed functions */
414: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetGlobal_C",NULL);
415: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetLocal_C",NULL);
416: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetCoarseLevel_C",NULL);
417: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetMatchOptimization_C",NULL);
418: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetBipart_C",NULL);
419: return(0);
420: }
422: /*MC
423: MATPARTITIONINGPARTY - Creates a partitioning context via the external package Party.
425: Level: beginner
427: Notes: See http://wwwcs.upb.de/fachbereich/AG/monien/RESEARCH/PART/party.html
429: .keywords: Partitioning, create, context
431: .seealso: MatPartitioningSetType(), MatPartitioningType
433: M*/
437: PETSC_EXTERN PetscErrorCode MatPartitioningCreate_Party(MatPartitioning part)
438: {
439: PetscErrorCode ierr;
440: MatPartitioning_Party *party;
443: PetscNewLog(part,MatPartitioning_Party,&party);
444: part->data = (void*)party;
446: PetscStrcpy(party->global,"gcf,gbf");
447: PetscStrcpy(party->local,"kl");
449: party->redm = PETSC_TRUE;
450: party->redo = PETSC_TRUE;
451: party->recursive = PETSC_TRUE;
452: party->verbose = PETSC_FALSE;
453: party->nbvtxcoarsed = 200;
455: part->ops->apply = MatPartitioningApply_Party;
456: part->ops->view = MatPartitioningView_Party;
457: part->ops->destroy = MatPartitioningDestroy_Party;
458: part->ops->setfromoptions = MatPartitioningSetFromOptions_Party;
460: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetGlobal_C",MatPartitioningPartySetGlobal_Party);
461: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetLocal_C",MatPartitioningPartySetLocal_Party);
462: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetCoarseLevel_C",MatPartitioningPartySetCoarseLevel_Party);
463: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetMatchOptimization_C",MatPartitioningPartySetMatchOptimization_Party);
464: PetscObjectComposeFunction((PetscObject)part,"MatPartitioningPartySetBipart_C",MatPartitioningPartySetBipart_Party);
465: return(0);
466: }