/[escript]/trunk/paso/src/AMG.c
ViewVC logotype

Annotation of /trunk/paso/src/AMG.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3259 - (hide annotations)
Mon Oct 11 01:48:14 2010 UTC (8 years, 11 months ago) by jfenwick
File MIME type: text/plain
File size: 19866 byte(s)
Merging dudley and scons updates from branches

1 gross 3193
2     /*******************************************************
3     *
4     * Copyright (c) 2003-2010 by University of Queensland
5     * Earth Systems Science Computational Center (ESSCC)
6     * http://www.uq.edu.au/esscc
7     *
8     * Primary Business: Queensland, Australia
9     * Licensed under the Open Software License version 3.0
10     * http://www.opensource.org/licenses/osl-3.0.php
11     *
12     *******************************************************/
13    
14    
15     /**************************************************************/
16    
17     /* Paso: AMG preconditioner */
18    
19     /**************************************************************/
20    
21     /* Author: artak@uq.edu.au */
22    
23     /**************************************************************/
24    
25     #include "Paso.h"
26     #include "Preconditioner.h"
27     #include "Options.h"
28     #include "PasoUtil.h"
29     #include "UMFPACK.h"
30     #include "MKL.h"
31     #include "Coarsening.h"
32    
33     /**************************************************************/
34    
35     /* free all memory used by AMG */
36    
37     void Paso_Preconditioner_LocalAMG_free(Paso_Preconditioner_LocalAMG * in) {
38     if (in!=NULL) {
39     Paso_Preconditioner_LocalSmoother_free(in->Smoother);
40    
41     /*=========================*/
42     Paso_SparseMatrix_free(in->A_FC);
43     Paso_SparseMatrix_free(in->A_FF);
44     Paso_SparseMatrix_free(in->W_FC);
45     Paso_SparseMatrix_free(in->A_CF);
46     Paso_SparseMatrix_free(in->P);
47     Paso_SparseMatrix_free(in->R);
48     Paso_SparseMatrix_free(in->A);
49     if(in->coarsest_level==TRUE) {
50     #ifdef MKL
51     Paso_MKL_free1(in->AOffset1);
52     Paso_SparseMatrix_free(in->AOffset1);
53     Paso_SparseMatrix_free(in->AUnrolled);
54     #else
55     #ifdef UMFPACK
56     Paso_UMFPACK1_free((Paso_UMFPACK_Handler*)(in->solver));
57     Paso_SparseMatrix_free(in->AUnrolled);
58     #endif
59     #endif
60     }
61     MEMFREE(in->rows_in_F);
62     MEMFREE(in->rows_in_C);
63     MEMFREE(in->mask_F);
64     MEMFREE(in->mask_C);
65     MEMFREE(in->x_F);
66     MEMFREE(in->b_F);
67     MEMFREE(in->x_C);
68     MEMFREE(in->b_C);
69     in->solver=NULL;
70     Paso_Preconditioner_LocalAMG_free(in->AMG_of_Coarse);
71     MEMFREE(in->b_C);
72     MEMFREE(in);
73     }
74     }
75    
76     /**************************************************************/
77    
78     /* constructs the block-block factorization of
79    
80     [ A_FF A_FC ]
81     A_p=
82     [ A_CF A_FF ]
83    
84     to
85    
86     [ I 0 ] [ A_FF 0 ] [ I invA_FF*A_FF ]
87     [ A_CF*invA_FF I ] [ 0 S ] [ 0 I ]
88    
89    
90     where S=A_FF-ACF*invA_FF*A_FC within the shape of S
91    
92     then AMG is applied to S again until S becomes empty
93    
94     */
95     Paso_Preconditioner_LocalAMG* Paso_Preconditioner_LocalAMG_alloc(Paso_SparseMatrix *A_p,dim_t level,Paso_Options* options) {
96     Paso_Preconditioner_LocalAMG* out=NULL;
97     /*
98     Paso_Pattern* temp1=NULL;
99     Paso_Pattern* temp2=NULL;
100     */
101     bool_t verbose=options->verbose;
102     bool_t timing=0;
103    
104     const dim_t n=A_p->numRows;
105     const dim_t n_block=A_p->row_block_size;
106    
107    
108     index_t* mis_marker=NULL;
109     index_t* counter=NULL;
110     /*index_t iPtr,*index, *where_p;*/
111     dim_t i,j;
112     Paso_SparseMatrix * A_c=NULL;
113     double time0=0;
114     Paso_SparseMatrix * Atemp=NULL;
115     double sparsity=0;
116    
117     /*
118     double *temp,*temp_1;
119     double S;
120     index_t iptr;
121     */
122    
123     /*char filename[8];*/
124    
125     /*
126     sprintf(filename,"AMGLevel%d",level);
127    
128     Paso_SparseMatrix_saveMM(A_p,filename);
129     */
130    
131     /*Make sure we have block sizes 1*/
132     /*if (A_p->col_block_size>1) {
133     Paso_setError(TYPE_ERROR,"Paso_Solver_getAMG: AMG requires column block size 1.");
134     return NULL;
135     }
136     if (A_p->row_block_size>1) {
137     Paso_setError(TYPE_ERROR,"Paso_Solver_getAMG: AMG requires row block size 1.");
138     return NULL;
139     }*/
140     out=MEMALLOC(1,Paso_Preconditioner_LocalAMG);
141    
142    
143     /* identify independend set of rows/columns */
144     mis_marker=TMPMEMALLOC(n,index_t);
145     counter=TMPMEMALLOC(n,index_t);
146 jfenwick 3259 if ( !( Esys_checkPtr(mis_marker) || Esys_checkPtr(counter) || Esys_checkPtr(out)) ) {
147 gross 3193
148    
149     out->post_sweeps=options->post_sweeps;
150     out->pre_sweeps=options->pre_sweeps;
151    
152     out->AMG_of_Coarse=NULL;
153     out->A_FF=NULL;
154     out->A_FC=NULL;
155     out->A_CF=NULL;
156     out->W_FC=NULL;
157     out->P=NULL;
158     out->R=NULL;
159     out->rows_in_F=NULL;
160     out->rows_in_C=NULL;
161     out->mask_F=NULL;
162     out->mask_C=NULL;
163     out->x_F=NULL;
164     out->b_F=NULL;
165     out->x_C=NULL;
166     out->b_C=NULL;
167     out->A=Paso_SparseMatrix_getReference(A_p);
168     out->AUnrolled=NULL;
169     out->AOffset1=NULL;
170     out->solver=NULL;
171     out->Smoother=NULL;
172     out->level=level;
173     out->n=n;
174     out->n_F=n+1;
175     out->n_block=n_block;
176    
177    
178     sparsity=(A_p->len*1.)/(1.*A_p->numRows*A_p->numCols);
179    
180     if (verbose) fprintf(stdout,"Stats: Sparsity of the Coarse Matrix with %d non-zeros (%d,%d) in level %d is %.6f\n",A_p->len,A_p->numRows,A_p->numCols,level,sparsity);
181    
182    
183     if(sparsity>0.05) {
184     level=0;
185     }
186    
187    
188     if (level==0 || n<=options->min_coarse_matrix_size) {
189     out->coarsest_level=TRUE;
190     #ifdef MKL
191     out->AUnrolled=Paso_SparseMatrix_unroll(A_p);
192     out->AOffset1=Paso_SparseMatrix_alloc(MATRIX_FORMAT_BLK1 + MATRIX_FORMAT_OFFSET1, out->AUnrolled->pattern,1,1, FALSE);
193     #pragma omp parallel for private(i) schedule(static)
194     for (i=0;i<out->A->len;++i) {
195     out->AOffset1->val[i]=out->AUnrolled->val[i];
196     }
197     #else
198     #ifdef UMFPACK
199     out->AUnrolled=Paso_SparseMatrix_unroll(A_p);
200     /*Paso_SparseMatrix_saveMM(out->AUnrolled,"AUnroll.mat");
201     Paso_SparseMatrix_saveMM(A_p,"Aorg.mat");
202     */
203     #else
204     out->Smoother=Paso_Preconditioner_LocalSmoother_alloc(A_p, (options->smoother == PASO_JACOBI), verbose);
205     #endif
206     #endif
207    
208     } else {
209     out->coarsest_level=FALSE;
210     out->Smoother=Paso_Preconditioner_LocalSmoother_alloc(A_p, (options->smoother == PASO_JACOBI), verbose);
211    
212     /* Start Coarsening : */
213 jfenwick 3259 time0=Esys_timer();
214 gross 3193 Paso_Coarsening_Local(mis_marker, A_p, options->coarsening_threshold, options->coarsening_method);
215 jfenwick 3259 time0=Esys_timer()-time0;
216 gross 3193 if (timing) fprintf(stdout,"Level %d: timing: Coarsening: %e\n",level,time0);
217    
218     #pragma omp parallel for private(i) schedule(static)
219     for (i = 0; i < n; ++i) {
220     mis_marker[i]=(mis_marker[i]== PASO_COARSENING_IN_F);
221     counter[i]=mis_marker[i];
222     }
223     out->n_F=Paso_Util_cumsum(n,counter);
224    
225     if (out->n_F==0) {
226     out->coarsest_level=TRUE;
227     level=0;
228     if (verbose) {
229     /*printf("AMG coarsening eliminates all unknowns, switching to Jacobi preconditioner.\n");*/
230     printf("AMG coarsening does not eliminate any of the unknowns, switching to Jacobi preconditioner.\n");
231     }
232     }
233     else if (out->n_F==n) {
234     out->coarsest_level=TRUE;
235     level=0;
236     if (verbose) {
237     /*printf("AMG coarsening eliminates all unknowns, switching to Jacobi preconditioner.\n");*/
238     printf("AMG coarsening eliminates all of the unknowns, switching to Jacobi preconditioner.\n");
239    
240     }
241     } else {
242    
243 jfenwick 3259 if (Esys_noError()) {
244 gross 3193
245     /*#pragma omp parallel for private(i) schedule(static)
246     for (i = 0; i < n; ++i) counter[i]=mis_marker[i];
247     out->n_F=Paso_Util_cumsum(n,counter);
248     */
249    
250     out->mask_F=MEMALLOC(n,index_t);
251     out->rows_in_F=MEMALLOC(out->n_F,index_t);
252 jfenwick 3259 if (! (Esys_checkPtr(out->mask_F) || Esys_checkPtr(out->rows_in_F) ) ) {
253 gross 3193 /* creates an index for F from mask */
254     #pragma omp parallel for private(i) schedule(static)
255     for (i = 0; i < out->n_F; ++i) out->rows_in_F[i]=-1;
256     #pragma omp parallel for private(i) schedule(static)
257     for (i = 0; i < n; ++i) {
258     if (mis_marker[i]) {
259     out->rows_in_F[counter[i]]=i;
260     out->mask_F[i]=counter[i];
261     } else {
262     out->mask_F[i]=-1;
263     }
264     }
265    
266     }
267     }
268    
269     /* if(level==1) {
270     printf("##TOTAL: %d, ELIMINATED: %d\n",n,out->n_F);
271     for (i = 0; i < n; ++i) {
272     printf("##%d %d\n",i,!mis_marker[i]);
273     }
274     }
275     */
276    
277     /*check whether coarsening process actually makes sense to continue.
278     if coarse matrix at least smaller by 30% then continue, otherwise we stop.*/
279     if ((out->n_F*100/n)<30) {
280     level=1;
281     }
282    
283 jfenwick 3259 if ( Esys_noError()) {
284 gross 3193 out->n_C=n-out->n_F;
285     out->rows_in_C=MEMALLOC(out->n_C,index_t);
286     out->mask_C=MEMALLOC(n,index_t);
287 jfenwick 3259 if (! (Esys_checkPtr(out->mask_C) || Esys_checkPtr(out->rows_in_C) ) ) {
288 gross 3193 /* creates an index for C from mask */
289     #pragma omp parallel for private(i) schedule(static)
290     for (i = 0; i < n; ++i) counter[i]=! mis_marker[i];
291     Paso_Util_cumsum(n,counter);
292     #pragma omp parallel for private(i) schedule(static)
293     for (i = 0; i < out->n_C; ++i) out->rows_in_C[i]=-1;
294     #pragma omp parallel for private(i) schedule(static)
295     for (i = 0; i < n; ++i) {
296     if (! mis_marker[i]) {
297     out->rows_in_C[counter[i]]=i;
298     out->mask_C[i]=counter[i];
299     } else {
300     out->mask_C[i]=-1;
301     }
302     }
303     }
304     }
305 jfenwick 3259 if ( Esys_noError()) {
306 gross 3193 /* get A_FF block: */
307     /*
308     out->A_FF=Paso_SparseMatrix_getSubmatrix(A_p,out->n_F,out->n_F,out->rows_in_F,out->mask_F);
309     out->A_CF=Paso_SparseMatrix_getSubmatrix(A_p,out->n_C,out->n_F,out->rows_in_C,out->mask_F);
310     out->A_FC=Paso_SparseMatrix_getSubmatrix(A_p,out->n_F,out->n_C,out->rows_in_F,out->mask_C);
311     */
312    
313     /*Compute W_FC*/
314     /*initialy W_FC=A_FC*/
315     out->W_FC=Paso_SparseMatrix_getSubmatrix(A_p,out->n_F,out->n_C,out->rows_in_F,out->mask_C);
316    
317     /*sprintf(filename,"W_FCbefore_%d",level);
318     Paso_SparseMatrix_saveMM(out->W_FC,filename);
319     */
320     /* for (i = 0; i < n; ++i) {
321     printf("##mis_marker[%d]=%d\n",i,mis_marker[i]);
322     }
323     */
324 jfenwick 3259 time0=Esys_timer();
325 gross 3193 Paso_SparseMatrix_updateWeights(A_p,out->W_FC,mis_marker);
326 jfenwick 3259 time0=Esys_timer()-time0;
327 gross 3193 if (timing) fprintf(stdout,"timing: updateWeights: %e\n",time0);
328    
329     /*
330     sprintf(filename,"W_FCafter_%d",level);
331     Paso_SparseMatrix_saveMM(out->W_FC,filename);
332     */
333    
334     /* get Prolongation and Restriction */
335 jfenwick 3259 time0=Esys_timer();
336 gross 3193 out->P=Paso_SparseMatrix_getProlongation(out->W_FC,mis_marker);
337 jfenwick 3259 time0=Esys_timer()-time0;
338 gross 3193 if (timing) fprintf(stdout,"timing: getProlongation: %e\n",time0);
339     /*out->P=Paso_SparseMatrix_loadMM_toCSR("P1.mtx");*/
340    
341     /*
342     sprintf(filename,"P_%d",level);
343     Paso_SparseMatrix_saveMM(out->P,filename);
344     */
345    
346 jfenwick 3259 time0=Esys_timer();
347 gross 3193 out->R=Paso_SparseMatrix_getRestriction(out->P);
348 jfenwick 3259 time0=Esys_timer()-time0;
349 gross 3193 if (timing) fprintf(stdout,"timing: getRestriction: %e\n",time0);
350     /*out->R=Paso_SparseMatrix_loadMM_toCSR("R1.mtx");*/
351    
352     /*
353     sprintf(filename,"R_%d",level);
354     Paso_SparseMatrix_saveMM(out->R,filename);
355     */
356    
357     }
358 jfenwick 3259 if ( Esys_noError()) {
359 gross 3193
360 jfenwick 3259 time0=Esys_timer();
361 gross 3193
362     Atemp=Paso_SparseMatrix_MatrixMatrix(A_p,out->P);
363    
364     A_c=Paso_SparseMatrix_MatrixMatrix(out->R,Atemp);
365    
366     /*A_c=Paso_SparseMatrix_loadMM_toCSR("A_C1.mtx");*/
367    
368     Paso_SparseMatrix_free(Atemp);
369    
370     /*A_c=Paso_Solver_getCoarseMatrix(A_p,out->R,out->P);*/
371 jfenwick 3259 time0=Esys_timer()-time0;
372 gross 3193 if (timing) fprintf(stdout,"timing: getCoarseMatrix: %e\n",time0);
373    
374    
375     /*Paso_Solver_getCoarseMatrix(A_c, A_p,out->R,out->P);*/
376     /*
377     sprintf(filename,"A_C_%d",level);
378     Paso_SparseMatrix_saveMM(A_c,filename);
379     */
380    
381     out->AMG_of_Coarse=Paso_Preconditioner_LocalAMG_alloc(A_c,level-1,options);
382     }
383    
384     /* allocate work arrays for AMG application */
385 jfenwick 3259 if (Esys_noError()) {
386 gross 3193 /*
387     out->x_F=MEMALLOC(n_block*out->n_F,double);
388     out->b_F=MEMALLOC(n_block*out->n_F,double);
389     */
390     out->x_C=MEMALLOC(n_block*out->n_C,double);
391     out->b_C=MEMALLOC(n_block*out->n_C,double);
392    
393 jfenwick 3259 /*if (! (Esys_checkPtr(out->x_F) || Esys_checkPtr(out->b_F) || Esys_checkPtr(out->x_C) || Esys_checkPtr(out->b_C) ) ) {*/
394     if ( ! ( Esys_checkPtr(out->x_C) || Esys_checkPtr(out->b_C) ) ) {
395 gross 3193
396     /*
397     #pragma omp parallel for private(i) schedule(static)
398     for (i = 0; i < out->n_F; ++i) {
399     out->x_F[i]=0.;
400     out->b_F[i]=0.;
401     }
402     */
403    
404     #pragma omp parallel for private(i,j) schedule(static)
405     for (i = 0; i < out->n_C; ++i) {
406     for(j=0;j<n_block;++j) {
407     out->x_C[i*n_block+j]=0.;
408     out->b_C[i*n_block+j]=0.;
409     }
410     }
411     }
412     }
413     Paso_SparseMatrix_free(A_c);
414     }
415     }
416     }
417     TMPMEMFREE(mis_marker);
418     TMPMEMFREE(counter);
419    
420 jfenwick 3259 if (Esys_noError()) {
421 gross 3193 if (verbose && level>0 && !out->coarsest_level) {
422     printf("AMG: level: %d: %d unknowns eliminated. %d left.\n",level, out->n_F,out->n_C);
423     }
424     return out;
425     } else {
426     Paso_Preconditioner_LocalAMG_free(out);
427     return NULL;
428     }
429     }
430    
431     /**************************************************************/
432    
433     /* apply AMG precondition b-> x
434    
435     in fact it solves
436    
437     [ I 0 ] [ A_FF 0 ] [ I invA_FF*A_FC ] [ x_F ] = [b_F]
438     [ A_CF*invA_FF I ] [ 0 S ] [ 0 I ] [ x_C ] = [b_C]
439    
440     in the form
441    
442     b->[b_F,b_C]
443     x_F=invA_FF*b_F
444     b_C=b_C-A_CF*x_F
445     x_C=AMG(b_C)
446     b_F=b_F-A_FC*x_C
447     x_F=invA_FF*b_F
448     x<-[x_F,x_C]
449    
450     should be called within a parallel region
451     barrier synconization should be performed to make sure that the input vector available
452    
453     */
454    
455     void Paso_Preconditioner_LocalAMG_solve(Paso_Preconditioner_LocalAMG * amg, double * x, double * b) {
456     dim_t i,j;
457     double time0=0;
458     double *r=NULL, *x0=NULL;
459     bool_t timing=0;
460    
461     dim_t post_sweeps=amg->post_sweeps;
462     dim_t pre_sweeps=amg->pre_sweeps;
463    
464     #ifdef UMFPACK
465     Paso_UMFPACK_Handler * ptr=NULL;
466     #endif
467    
468     r=MEMALLOC(amg->n*amg->n_block,double);
469     x0=MEMALLOC(amg->n*amg->n_block,double);
470    
471     if (amg->coarsest_level) {
472    
473 jfenwick 3259 time0=Esys_timer();
474 gross 3193 /*If all unknown are eliminated then Jacobi is the best preconditioner*/
475    
476     if (amg->n_F==0 || amg->n_F==amg->n) {
477     Paso_Preconditioner_LocalSmoother_solve(amg->A, amg->Smoother,x,b,1, FALSE);
478     }
479     else {
480     #ifdef MKL
481     Paso_MKL1(amg->AOffset1,x,b,timing);
482     #else
483     #ifdef UMFPACK
484     ptr=(Paso_UMFPACK_Handler *)(amg->solver);
485     Paso_UMFPACK1(&ptr,amg->AUnrolled,x,b,timing);
486     amg->solver=(void*) ptr;
487     #else
488     Paso_Preconditioner_LocalSmoother_solve(amg->A, amg->Smoother,x,b,1, FALSE);
489     #endif
490     #endif
491     }
492    
493 jfenwick 3259 time0=Esys_timer()-time0;
494 gross 3193 if (timing) fprintf(stdout,"timing: DIRECT SOLVER: %e\n",time0);
495    
496     } else {
497     /* presmoothing */
498 jfenwick 3259 time0=Esys_timer();
499 gross 3193 Paso_Preconditioner_LocalSmoother_solve(amg->A, amg->Smoother, x, b, pre_sweeps, FALSE);
500 jfenwick 3259 time0=Esys_timer()-time0;
501 gross 3193 if (timing) fprintf(stdout,"timing: Presmooting: %e\n",time0);
502    
503     /* end of presmoothing */
504    
505 jfenwick 3259 time0=Esys_timer();
506 gross 3193 #pragma omp parallel for private(i,j) schedule(static)
507     for (i=0;i<amg->n;++i) {
508     for (j=0;j<amg->n_block;++j) {
509     r[i*amg->n_block+j]=b[i*amg->n_block+j];
510     }
511     }
512    
513     /*r=b-Ax*/
514     Paso_SparseMatrix_MatrixVector_CSR_OFFSET0(-1.,amg->A,x,1.,r);
515    
516     /* b_c = R*r */
517     Paso_SparseMatrix_MatrixVector_CSR_OFFSET0(1.,amg->R,r,0.,amg->b_C);
518    
519 jfenwick 3259 time0=Esys_timer()-time0;
520 gross 3193 if (timing) fprintf(stdout,"timing: Before next level: %e\n",time0);
521    
522     /* x_C=AMG(b_C) */
523     Paso_Preconditioner_LocalAMG_solve(amg->AMG_of_Coarse,amg->x_C,amg->b_C);
524    
525 jfenwick 3259 time0=Esys_timer();
526 gross 3193
527     /* x_0 = P*x_c */
528     Paso_SparseMatrix_MatrixVector_CSR_OFFSET0(1.,amg->P,amg->x_C,0.,x0);
529    
530     /* x=x+x0 */
531     #pragma omp parallel for private(i,j) schedule(static)
532     for (i=0;i<amg->n;++i) {
533     for (j=0;j<amg->n_block;++j) {
534     x[i*amg->n_block+j]+=x0[i*amg->n_block+j];
535     }
536     }
537    
538     /*postsmoothing*/
539    
540     /*solve Ax=b with initial guess x */
541 jfenwick 3259 time0=Esys_timer();
542 gross 3193 Paso_Preconditioner_LocalSmoother_solve(amg->A, amg->Smoother, x, b, post_sweeps, TRUE);
543 jfenwick 3259 time0=Esys_timer()-time0;
544 gross 3193 if (timing) fprintf(stdout,"timing: Postsmoothing: %e\n",time0);
545    
546     /*end of postsmoothing*/
547    
548     }
549     MEMFREE(r);
550     MEMFREE(x0);
551    
552     return;
553     }

  ViewVC Help
Powered by ViewVC 1.1.26