Actual source code: nn.c
1: /*$Id: nn.c,v 1.13 2001/08/07 03:03:41 balay Exp $*/
3: #include src/sles/pc/impls/is/nn/nn.h
5: /* -------------------------------------------------------------------------- */
6: /*
7: PCSetUp_NN - Prepares for the use of the NN preconditioner
8: by setting data structures and options.
10: Input Parameter:
11: . pc - the preconditioner context
13: Application Interface Routine: PCSetUp()
15: Notes:
16: The interface routine PCSetUp() is not usually called directly by
17: the user, but instead is called by PCApply() if necessary.
18: */
19: static int PCSetUp_NN(PC pc)
20: {
22:
25: if (pc->setupcalled == 0) {
26: /* Set up all the "iterative substructuring" common block */
27: PCISSetUp(pc);
28: /* Create the coarse matrix. */
29: PCNNCreateCoarseMatrix(pc);
30: }
32: return(0);
33: }
35: /* -------------------------------------------------------------------------- */
36: /*
37: PCApply_NN - Applies the NN preconditioner to a vector.
39: Input Parameters:
40: . pc - the preconditioner context
41: . r - input vector (global)
43: Output Parameter:
44: . z - output vector (global)
46: Application Interface Routine: PCApply()
47: */
48: static int PCApply_NN(PC pc,Vec r,Vec z)
49: {
50: PC_IS *pcis = (PC_IS*)(pc->data);
51: int ierr,its;
52: PetscScalar m_one = -1.0;
53: Vec w = pcis->vec1_global;
57: /*
58: Dirichlet solvers.
59: Solving $ B_I^{(i)}r_I^{(i)} $ at each processor.
60: Storing the local results at vec2_D
61: */
62: VecScatterBegin(r,pcis->vec1_D,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_D);
63: VecScatterEnd (r,pcis->vec1_D,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_D);
64: SLESSolve(pcis->sles_D,pcis->vec1_D,pcis->vec2_D,&its);
65:
66: /*
67: Computing $ r_B - sum_j tilde R_j^T A_{BI}^{(j)} (B_I^{(j)}r_I^{(j)}) $ .
68: Storing the result in the interface portion of the global vector w.
69: */
70: MatMult(pcis->A_BI,pcis->vec2_D,pcis->vec1_B);
71: VecScale(&m_one,pcis->vec1_B);
72: VecCopy(r,w);
73: VecScatterBegin(pcis->vec1_B,w,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
74: VecScatterEnd (pcis->vec1_B,w,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
76: /*
77: Apply the interface preconditioner
78: */
79: PCNNApplyInterfacePreconditioner(pc,w,z,pcis->work_N,pcis->vec1_B,pcis->vec2_B,pcis->vec3_B,pcis->vec1_D,
80: pcis->vec3_D,pcis->vec1_N,pcis->vec2_N);
82: /*
83: Computing $ t_I^{(i)} = A_{IB}^{(i)} tilde R_i z_B $
84: The result is stored in vec1_D.
85: */
86: VecScatterBegin(z,pcis->vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
87: VecScatterEnd (z,pcis->vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
88: MatMult(pcis->A_IB,pcis->vec1_B,pcis->vec1_D);
90: /*
91: Dirichlet solvers.
92: Computing $ B_I^{(i)}t_I^{(i)} $ and sticking into the global vector the blocks
93: $ B_I^{(i)}r_I^{(i)} - B_I^{(i)}t_I^{(i)} $.
94: */
95: VecScatterBegin(pcis->vec2_D,z,INSERT_VALUES,SCATTER_REVERSE,pcis->global_to_D);
96: VecScatterEnd (pcis->vec2_D,z,INSERT_VALUES,SCATTER_REVERSE,pcis->global_to_D);
97: SLESSolve(pcis->sles_D,pcis->vec1_D,pcis->vec2_D,&its);
98: VecScale(&m_one,pcis->vec2_D);
99: VecScatterBegin(pcis->vec2_D,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_D);
100: VecScatterEnd (pcis->vec2_D,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_D);
102: return(0);
103: }
105: /* -------------------------------------------------------------------------- */
106: /*
107: PCDestroy_NN - Destroys the private context for the NN preconditioner
108: that was created with PCCreate_NN().
110: Input Parameter:
111: . pc - the preconditioner context
113: Application Interface Routine: PCDestroy()
114: */
115: static int PCDestroy_NN(PC pc)
116: {
117: PC_NN *pcnn = (PC_NN*)pc->data;
118: int ierr;
122: PCISDestroy(pc);
124: if (pcnn->coarse_mat) {MatDestroy(pcnn->coarse_mat);}
125: if (pcnn->coarse_x) {VecDestroy(pcnn->coarse_x);}
126: if (pcnn->coarse_b) {VecDestroy(pcnn->coarse_b);}
127: if (pcnn->sles_coarse) {SLESDestroy(pcnn->sles_coarse);}
128: if (pcnn->DZ_IN) {
129: if (pcnn->DZ_IN[0]) {PetscFree(pcnn->DZ_IN[0]);}
130: PetscFree(pcnn->DZ_IN);
131: }
133: /*
134: Free the private data structure that was hanging off the PC
135: */
136: PetscFree(pcnn);
137: return(0);
138: }
140: /* -------------------------------------------------------------------------- */
141: /*
142: PCCreate_NN - Creates a NN preconditioner context, PC_NN,
143: and sets this as the private data within the generic preconditioning
144: context, PC, that was created within PCCreate().
146: Input Parameter:
147: . pc - the preconditioner context
149: Application Interface Routine: PCCreate()
150: */
151: EXTERN_C_BEGIN
152: int PCCreate_NN(PC pc)
153: {
155: PC_NN *pcnn;
159: /*
160: Creates the private data structure for this preconditioner and
161: attach it to the PC object.
162: */
163: ierr = PetscNew(PC_NN,&pcnn);
164: pc->data = (void*)pcnn;
166: /*
167: Logs the memory usage; this is not needed but allows PETSc to
168: monitor how much memory is being used for various purposes.
169: */
170: PetscLogObjectMemory(pc,sizeof(PC_NN)+sizeof(PC_IS)); /* Is this the right thing to do? */
172: PCISCreate(pc);
173: pcnn->coarse_mat = 0;
174: pcnn->coarse_x = 0;
175: pcnn->coarse_b = 0;
176: pcnn->sles_coarse = 0;
177: pcnn->DZ_IN = 0;
179: /*
180: Set the pointers for the functions that are provided above.
181: Now when the user-level routines (such as PCApply(), PCDestroy(), etc.)
182: are called, they will automatically call these functions. Note we
183: choose not to provide a couple of these functions since they are
184: not needed.
185: */
186: pc->ops->apply = PCApply_NN;
187: pc->ops->applytranspose = 0;
188: pc->ops->setup = PCSetUp_NN;
189: pc->ops->destroy = PCDestroy_NN;
190: pc->ops->view = 0;
191: pc->ops->applyrichardson = 0;
192: pc->ops->applysymmetricleft = 0;
193: pc->ops->applysymmetricright = 0;
195: return(0);
196: }
197: EXTERN_C_END
200: /* -------------------------------------------------------------------------- */
201: /*
202: PCNNCreateCoarseMatrix -
203: */
204: int PCNNCreateCoarseMatrix (PC pc)
205: {
206: MPI_Request *send_request, *recv_request;
207: int i, j, k, ierr;
209: PetscScalar* mat; /* Sub-matrix with this subdomain's contribution to the coarse matrix */
210: PetscScalar** DZ_OUT; /* proc[k].DZ_OUT[i][] = bit of vector to be sent from processor k to processor i */
212: /* aliasing some names */
213: PC_IS* pcis = (PC_IS*)(pc->data);
214: PC_NN* pcnn = (PC_NN*)pc->data;
215: int n_neigh = pcis->n_neigh;
216: int* neigh = pcis->neigh;
217: int* n_shared = pcis->n_shared;
218: int** shared = pcis->shared;
219: PetscScalar** DZ_IN; /* Must be initialized after memory allocation. */
223: /* Allocate memory for mat (the +1 is to handle the case n_neigh equal to zero) */
224: PetscMalloc((n_neigh*n_neigh+1)*sizeof(PetscScalar),&mat);
226: /* Allocate memory for DZ */
227: /* Notice that DZ_OUT[0] is allocated some space that is never used. */
228: /* This is just in order to DZ_OUT and DZ_IN to have exactly the same form. */
229: {
230: int size_of_Z = 0;
231: ierr = PetscMalloc ((n_neigh+1)*sizeof(PetscScalar*),&pcnn->DZ_IN);
232: DZ_IN = pcnn->DZ_IN;
233: ierr = PetscMalloc ((n_neigh+1)*sizeof(PetscScalar*),&DZ_OUT);
234: for (i=0; i<n_neigh; i++) {
235: size_of_Z += n_shared[i];
236: }
237: PetscMalloc ((size_of_Z+1)*sizeof(PetscScalar),&DZ_IN[0]);
238: PetscMalloc ((size_of_Z+1)*sizeof(PetscScalar),&DZ_OUT[0]);
239: }
240: for (i=1; i<n_neigh; i++) {
241: DZ_IN[i] = DZ_IN [i-1] + n_shared[i-1];
242: DZ_OUT[i] = DZ_OUT[i-1] + n_shared[i-1];
243: }
245: /* Set the values of DZ_OUT, in order to send this info to the neighbours */
246: /* First, set the auxiliary array pcis->work_N. */
247: PCISScatterArrayNToVecB(pcis->work_N,pcis->D,INSERT_VALUES,SCATTER_REVERSE,pc);
248: for (i=1; i<n_neigh; i++){
249: for (j=0; j<n_shared[i]; j++) {
250: DZ_OUT[i][j] = pcis->work_N[shared[i][j]];
251: }
252: }
254: /* Non-blocking send/receive the common-interface chunks of scaled nullspaces */
255: /* Notice that send_request[] and recv_request[] could have one less element. */
256: /* We make them longer to have request[i] corresponding to neigh[i]. */
257: {
258: int tag;
259: PetscObjectGetNewTag((PetscObject)pc,&tag);
260: PetscMalloc((2*(n_neigh)+1)*sizeof(MPI_Request),&send_request);
261: recv_request = send_request + (n_neigh);
262: for (i=1; i<n_neigh; i++) {
263: MPI_Isend((void*)(DZ_OUT[i]),n_shared[i],MPIU_SCALAR,neigh[i],tag,pc->comm,&(send_request[i]));
264: MPI_Irecv((void*)(DZ_IN [i]),n_shared[i],MPIU_SCALAR,neigh[i],tag,pc->comm,&(recv_request[i]));
265: }
266: }
268: /* Set DZ_IN[0][] (recall that neigh[0]==rank, always) */
269: for(j=0; j<n_shared[0]; j++) {
270: DZ_IN[0][j] = pcis->work_N[shared[0][j]];
271: }
273: /* Start computing with local D*Z while communication goes on. */
274: /* Apply Schur complement. The result is "stored" in vec (more */
275: /* precisely, vec points to the result, stored in pc_nn->vec1_B) */
276: /* and also scattered to pcnn->work_N. */
277: PCNNApplySchurToChunk(pc,n_shared[0],shared[0],DZ_IN[0],pcis->work_N,pcis->vec1_B,
278: pcis->vec2_B,pcis->vec1_D,pcis->vec2_D);
280: /* Compute the first column, while completing the receiving. */
281: for (i=0; i<n_neigh; i++) {
282: MPI_Status stat;
283: int ind=0;
284: if (i>0) { MPI_Waitany(n_neigh-1,recv_request+1,&ind,&stat); ind++;}
285: mat[ind*n_neigh+0] = 0.0;
286: for (k=0; k<n_shared[ind]; k++) {
287: mat[ind*n_neigh+0] += DZ_IN[ind][k] * pcis->work_N[shared[ind][k]];
288: }
289: }
291: /* Compute the remaining of the columns */
292: for (j=1; j<n_neigh; j++) {
293: PCNNApplySchurToChunk(pc,n_shared[j],shared[j],DZ_IN[j],pcis->work_N,pcis->vec1_B,
294: pcis->vec2_B,pcis->vec1_D,pcis->vec2_D);
295: for (i=0; i<n_neigh; i++) {
296: mat[i*n_neigh+j] = 0.0;
297: for (k=0; k<n_shared[i]; k++) {
298: mat[i*n_neigh+j] += DZ_IN[i][k] * pcis->work_N[shared[i][k]];
299: }
300: }
301: }
303: /* Complete the sending. */
304: if (n_neigh>1) {
305: MPI_Status *stat;
306: PetscMalloc((n_neigh-1)*sizeof(MPI_Status),&stat);
307: MPI_Waitall(n_neigh-1,&(send_request[1]),stat);
308: PetscFree(stat);
309: }
311: /* Free the memory for the MPI requests */
312: PetscFree(send_request);
314: /* Free the memory for DZ_OUT */
315: if (DZ_OUT) {
316: if (DZ_OUT[0]) { PetscFree(DZ_OUT[0]); }
317: PetscFree(DZ_OUT);
318: }
320: {
321: int size,n_neigh_m1;
322: MPI_Comm_size(pc->comm,&size);
323: n_neigh_m1 = (n_neigh) ? n_neigh-1 : 0;
324: /* Create the global coarse vectors (rhs and solution). */
325: VecCreateMPI(pc->comm,1,size,&(pcnn->coarse_b));
326: VecDuplicate(pcnn->coarse_b,&(pcnn->coarse_x));
327: /* Create and set the global coarse matrix. */
328: MatCreateMPIAIJ(pc->comm,1,1,size,size,1,PETSC_NULL,n_neigh_m1,PETSC_NULL,&(pcnn->coarse_mat));
329: MatSetValues(pcnn->coarse_mat,n_neigh,neigh,n_neigh,neigh,mat,ADD_VALUES);
330: MatAssemblyBegin(pcnn->coarse_mat,MAT_FINAL_ASSEMBLY);
331: MatAssemblyEnd (pcnn->coarse_mat,MAT_FINAL_ASSEMBLY);
332: }
334: {
335: int rank;
336: PetscScalar one = 1.0;
337: IS is;
338: MPI_Comm_rank(pc->comm,&rank);
339: /* "Zero out" rows of not-purely-Neumann subdomains */
340: if (pcis->pure_neumann) { /* does NOT zero the row; create an empty index set. The reason is that MatZeroRows() is collective. */
341: ISCreateStride(pc->comm,0,0,0,&is);
342: } else { /* here it DOES zero the row, since it's not a floating subdomain. */
343: ISCreateStride(pc->comm,1,rank,0,&is);
344: }
345: MatZeroRows(pcnn->coarse_mat,is,&one);
346: ISDestroy(is);
347: }
349: /* Create the coarse linear solver context */
350: {
351: PC pc_ctx, inner_pc;
352: KSP ksp_ctx;
353: SLESCreate(pc->comm,&pcnn->sles_coarse);
354: SLESSetOperators(pcnn->sles_coarse,pcnn->coarse_mat,pcnn->coarse_mat,SAME_PRECONDITIONER);
355: SLESGetKSP(pcnn->sles_coarse,&ksp_ctx);
356: SLESGetPC(pcnn->sles_coarse,&pc_ctx);
357: PCSetType(pc_ctx,PCREDUNDANT);
358: KSPSetType(ksp_ctx,KSPPREONLY);
359: PCRedundantGetPC(pc_ctx,&inner_pc);
360: PCSetType(inner_pc,PCLU);
361: SLESSetOptionsPrefix(pcnn->sles_coarse,"coarse_");
362: SLESSetFromOptions(pcnn->sles_coarse);
363: /* the vectors in the following line are dummy arguments, just telling the SLES the vector size. Values are not used */
364: SLESSetUp(pcnn->sles_coarse,pcnn->coarse_x,pcnn->coarse_b);
365: }
367: /* Free the memory for mat */
368: PetscFree(mat);
370: /* for DEBUGGING, save the coarse matrix to a file. */
371: {
372: PetscTruth flg;
373: PetscOptionsHasName(PETSC_NULL,"-save_coarse_matrix",&flg);
374: if (flg) {
375: PetscViewer viewer;
376: PetscViewerASCIIOpen(PETSC_COMM_WORLD,"coarse.m",&viewer);
377: PetscViewerSetFormat(viewer,PETSC_VIEWER_ASCII_MATLAB);
378: MatView(pcnn->coarse_mat,viewer);
379: PetscViewerDestroy(viewer);
380: }
381: }
383: /* Set the variable pcnn->factor_coarse_rhs. */
384: pcnn->factor_coarse_rhs = (pcis->pure_neumann) ? 1.0 : 0.0;
386: /* See historical note 02, at the bottom of this file. */
388: return(0);
389: }
391: /* -------------------------------------------------------------------------- */
392: /*
393: PCNNApplySchurToChunk -
395: Input parameters:
396: . pcnn
397: . n - size of chunk
398: . idx - indices of chunk
399: . chunk - values
401: Output parameters:
402: . array_N - result of Schur complement applied to chunk, scattered to big array
403: . vec1_B - result of Schur complement applied to chunk
404: . vec2_B - garbage (used as work space)
405: . vec1_D - garbage (used as work space)
406: . vec2_D - garbage (used as work space)
408: */
409: int PCNNApplySchurToChunk(PC pc, int n, int* idx, PetscScalar *chunk, PetscScalar* array_N, Vec vec1_B, Vec vec2_B, Vec vec1_D, Vec vec2_D)
410: {
411: int i, ierr;
413: PC_IS *pcis = (PC_IS*)(pc->data);
417: PetscMemzero((void*)array_N, pcis->n*sizeof(PetscScalar));
418: for (i=0; i<n; i++) { array_N[idx[i]] = chunk[i]; }
419: PCISScatterArrayNToVecB(array_N,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pc);
420: PCISApplySchur(pc,vec2_B,vec1_B,(Vec)0,vec1_D,vec2_D);
421: PCISScatterArrayNToVecB(array_N,vec1_B,INSERT_VALUES,SCATTER_REVERSE,pc);
423: return(0);
424: }
426: /* -------------------------------------------------------------------------- */
427: /*
428: PCNNApplyInterfacePreconditioner - Apply the interface preconditioner, i.e.,
429: the preconditioner for the Schur complement.
431: Input parameter:
432: . r - global vector of interior and interface nodes. The values on the interior nodes are NOT used.
434: Output parameters:
435: . z - global vector of interior and interface nodes. The values on the interface are the result of
436: the application of the interface preconditioner to the interface part of r. The values on the
437: interior nodes are garbage.
438: . work_N - array of local nodes (interior and interface, including ghosts); returns garbage (used as work space)
439: . vec1_B - vector of local interface nodes (including ghosts); returns garbage (used as work space)
440: . vec2_B - vector of local interface nodes (including ghosts); returns garbage (used as work space)
441: . vec3_B - vector of local interface nodes (including ghosts); returns garbage (used as work space)
442: . vec1_D - vector of local interior nodes; returns garbage (used as work space)
443: . vec2_D - vector of local interior nodes; returns garbage (used as work space)
444: . vec1_N - vector of local nodes (interior and interface, including ghosts); returns garbage (used as work space)
445: . vec2_N - vector of local nodes (interior and interface, including ghosts); returns garbage (used as work space)
447: */
448: int PCNNApplyInterfacePreconditioner (PC pc, Vec r, Vec z, PetscScalar* work_N, Vec vec1_B, Vec vec2_B, Vec vec3_B, Vec vec1_D,
449: Vec vec2_D, Vec vec1_N, Vec vec2_N)
450: {
453: PC_IS* pcis = (PC_IS*)(pc->data);
457: /*
458: First balancing step.
459: */
460: {
461: PetscTruth flg;
462: PetscOptionsHasName(PETSC_NULL,"-turn_off_first_balancing",&flg);
463: if (!flg) {
464: PCNNBalancing(pc,r,(Vec)0,z,vec1_B,vec2_B,(Vec)0,vec1_D,vec2_D,work_N);
465: } else {
466: VecCopy(r,z);
467: }
468: }
470: /*
471: Extract the local interface part of z and scale it by D
472: */
473: VecScatterBegin(z,vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
474: VecScatterEnd (z,vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
475: VecPointwiseMult(pcis->D,vec1_B,vec2_B);
477: /* Neumann Solver */
478: PCISApplyInvSchur(pc,vec2_B,vec1_B,vec1_N,vec2_N);
480: /*
481: Second balancing step.
482: */
483: {
484: PetscTruth flg;
485: PetscOptionsHasName(PETSC_NULL,"-turn_off_second_balancing",&flg);
486: if (!flg) {
487: PCNNBalancing(pc,r,vec1_B,z,vec2_B,vec3_B,(Vec)0,vec1_D,vec2_D,work_N);
488: } else {
489: PetscScalar zero = 0.0;
490: VecPointwiseMult(pcis->D,vec1_B,vec2_B);
491: VecSet(&zero,z);
492: VecScatterBegin(vec2_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
493: VecScatterEnd (vec2_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
494: }
495: }
497: return(0);
498: }
500: /* -------------------------------------------------------------------------- */
501: /*
502: PCNNBalancing - Computes z, as given in equations (15) and (16) (if the
503: input argument u is provided), or s, as given in equations
504: (12) and (13), if the input argument u is a null vector.
505: Notice that the input argument u plays the role of u_i in
506: equation (14). The equation numbers refer to [Man93].
508: Input Parameters:
509: . pcnn - NN preconditioner context.
510: . r - MPI vector of all nodes (interior and interface). It's preserved.
511: . u - (Optional) sequential vector of local interface nodes. It's preserved UNLESS vec3_B is null.
513: Output Parameters:
514: . z - MPI vector of interior and interface nodes. Returns s or z (see description above).
515: . vec1_B - Sequential vector of local interface nodes. Workspace.
516: . vec2_B - Sequential vector of local interface nodes. Workspace.
517: . vec3_B - (Optional) sequential vector of local interface nodes. Workspace.
518: . vec1_D - Sequential vector of local interior nodes. Workspace.
519: . vec2_D - Sequential vector of local interior nodes. Workspace.
520: . work_N - Array of all local nodes (interior and interface). Workspace.
522: */
523: int PCNNBalancing (PC pc, Vec r, Vec u, Vec z, Vec vec1_B, Vec vec2_B, Vec vec3_B,
524: Vec vec1_D, Vec vec2_D, PetscScalar *work_N)
525: {
526: int k, ierr, its;
527: PetscScalar zero = 0.0;
528: PetscScalar m_one = -1.0;
529: PetscScalar value;
530: PetscScalar* lambda;
531: PC_NN* pcnn = (PC_NN*)(pc->data);
532: PC_IS* pcis = (PC_IS*)(pc->data);
535: PetscLogEventBegin(PC_ApplyCoarse,0,0,0,0);
537: if (u) {
538: if (!vec3_B) { vec3_B = u; }
539: VecPointwiseMult(pcis->D,u,vec1_B);
540: VecSet(&zero,z);
541: VecScatterBegin(vec1_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
542: VecScatterEnd (vec1_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
543: VecScatterBegin(z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
544: VecScatterEnd (z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
545: PCISApplySchur(pc,vec2_B,vec3_B,(Vec)0,vec1_D,vec2_D);
546: VecScale(&m_one,vec3_B);
547: VecCopy(r,z);
548: VecScatterBegin(vec3_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
549: VecScatterEnd (vec3_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
550: } else {
551: VecCopy(r,z);
552: }
553: VecScatterBegin(z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
554: VecScatterEnd (z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
555: PCISScatterArrayNToVecB(work_N,vec2_B,INSERT_VALUES,SCATTER_REVERSE,pc);
556: for (k=0, value=0.0; k<pcis->n_shared[0]; k++) { value += pcnn->DZ_IN[0][k] * work_N[pcis->shared[0][k]]; }
557: value *= pcnn->factor_coarse_rhs; /* This factor is set in CreateCoarseMatrix(). */
558: {
559: int rank;
560: MPI_Comm_rank(pc->comm,&rank);
561: VecSetValue(pcnn->coarse_b,rank,value,INSERT_VALUES);
562: /*
563: Since we are only inserting local values (one value actually) we don't need to do the
564: reduction that tells us there is no data that needs to be moved. Hence we comment out these
565: VecAssemblyBegin(pcnn->coarse_b);
566: VecAssemblyEnd (pcnn->coarse_b);
567: */
568: }
569: SLESSolve(pcnn->sles_coarse,pcnn->coarse_b,pcnn->coarse_x,&its);
570: if (!u) { VecScale(&m_one,pcnn->coarse_x); }
571: VecGetArray(pcnn->coarse_x,&lambda);
572: for (k=0; k<pcis->n_shared[0]; k++) { work_N[pcis->shared[0][k]] = *lambda * pcnn->DZ_IN[0][k]; }
573: VecRestoreArray(pcnn->coarse_x,&lambda);
574: PCISScatterArrayNToVecB(work_N,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pc);
575: VecSet(&zero,z);
576: VecScatterBegin(vec2_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
577: VecScatterEnd (vec2_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
578: if (!u) {
579: VecScatterBegin(z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
580: VecScatterEnd (z,vec2_B,INSERT_VALUES,SCATTER_FORWARD,pcis->global_to_B);
581: PCISApplySchur(pc,vec2_B,vec1_B,(Vec)0,vec1_D,vec2_D);
582: VecCopy(r,z);
583: }
584: VecScatterBegin(vec1_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
585: VecScatterEnd (vec1_B,z,ADD_VALUES,SCATTER_REVERSE,pcis->global_to_B);
586: PetscLogEventEnd(PC_ApplyCoarse,0,0,0,0);
588: return(0);
589: }
594: /* ------- E N D O F T H E C O D E ------- */
595: /* */
596: /* From now on, "footnotes" (or "historical notes"). */
597: /* */
598: /* ------------------------------------------------- */
601: #ifdef __HISTORICAL_NOTES___do_not_compile__
603: /* --------------------------------------------------------------------------
604: Historical note 01
605: -------------------------------------------------------------------------- */
606: /*
607: We considered the possibility of an alternative D_i that would still
608: provide a partition of unity (i.e., $ sum_i N_i D_i N_i^T = I $).
609: The basic principle was still the pseudo-inverse of the counting
610: function; the difference was that we would not count subdomains
611: that do not contribute to the coarse space (i.e., not pure-Neumann
612: subdomains).
614: This turned out to be a bad idea: we would solve trivial Neumann
615: problems in the not pure-Neumann subdomains, since we would be scaling
616: the balanced residual by zero.
617: */
619: {
620: PetscTruth flg;
621: PetscOptionsHasName(PETSC_NULL,"-pcnn_new_scaling",&flg);
622: if (flg) {
623: Vec counter;
624: PetscScalar one=1.0, zero=0.0;
625: VecDuplicate(pc->vec,&counter);
626: VecSet(&zero,counter);
627: if (pcnn->pure_neumann) {
628: VecSet(&one,pcnn->D);
629: } else {
630: VecSet(&zero,pcnn->D);
631: }
632: VecScatterBegin(pcnn->D,counter,ADD_VALUES,SCATTER_REVERSE,pcnn->global_to_B);
633: VecScatterEnd (pcnn->D,counter,ADD_VALUES,SCATTER_REVERSE,pcnn->global_to_B);
634: VecScatterBegin(counter,pcnn->D,INSERT_VALUES,SCATTER_FORWARD,pcnn->global_to_B);
635: VecScatterEnd (counter,pcnn->D,INSERT_VALUES,SCATTER_FORWARD,pcnn->global_to_B);
636: VecDestroy(counter);
637: if (pcnn->pure_neumann) {
638: VecReciprocal(pcnn->D);
639: } else {
640: VecSet(&zero,pcnn->D);
641: }
642: }
643: }
647: /* --------------------------------------------------------------------------
648: Historical note 02
649: -------------------------------------------------------------------------- */
650: /*
651: We tried an alternative coarse problem, that would eliminate exactly a
652: constant error. Turned out not to improve the overall convergence.
653: */
655: /* Set the variable pcnn->factor_coarse_rhs. */
656: {
657: PetscTruth flg;
658: PetscOptionsHasName(PETSC_NULL,"-enforce_preserving_constants",&flg);
659: if (!flg) { pcnn->factor_coarse_rhs = (pcnn->pure_neumann) ? 1.0 : 0.0; }
660: else {
661: PetscScalar zero = 0.0, one = 1.0;
662: VecSet(&one,pcnn->vec1_B);
663: ApplySchurComplement(pcnn,pcnn->vec1_B,pcnn->vec2_B,(Vec)0,pcnn->vec1_D,pcnn->vec2_D);
664: VecSet(&zero,pcnn->vec1_global);
665: VecScatterBegin(pcnn->vec2_B,pcnn->vec1_global,ADD_VALUES,SCATTER_REVERSE,pcnn->global_to_B);
666: VecScatterEnd (pcnn->vec2_B,pcnn->vec1_global,ADD_VALUES,SCATTER_REVERSE,pcnn->global_to_B);
667: VecScatterBegin(pcnn->vec1_global,pcnn->vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcnn->global_to_B);
668: VecScatterEnd (pcnn->vec1_global,pcnn->vec1_B,INSERT_VALUES,SCATTER_FORWARD,pcnn->global_to_B);
669: if (pcnn->pure_neumann) { pcnn->factor_coarse_rhs = 1.0; }
670: else {
671: ScatterArrayNToVecB(pcnn->work_N,pcnn->vec1_B,INSERT_VALUES,SCATTER_REVERSE,pcnn);
672: for (k=0, pcnn->factor_coarse_rhs=0.0; k<pcnn->n_shared[0]; k++) {
673: pcnn->factor_coarse_rhs += pcnn->work_N[pcnn->shared[0][k]] * pcnn->DZ_IN[0][k];
674: }
675: if (pcnn->factor_coarse_rhs) { pcnn->factor_coarse_rhs = 1.0 / pcnn->factor_coarse_rhs; }
676: else { SETERRQ(1,"Constants cannot be preserved. Remove "-enforce_preserving_constants" option."); }
677: }
678: }
679: }
681: #endif /* __HISTORICAL_NOTES___do_not_compile */