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

Contents of /trunk/paso/src/SystemMatrix.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3369 - (show annotations)
Fri Nov 19 06:26:11 2010 UTC (8 years, 10 months ago) by gross
File MIME type: text/plain
File size: 12012 byte(s)
matrix balance introduced in paso
1
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: SystemMatrix */
18
19 /**************************************************************/
20
21 /* Author: Lutz Gross, l.gross@uq.edu.au */
22
23 /**************************************************************/
24
25 #include "SystemMatrix.h"
26 #include "Preconditioner.h"
27
28 /**************************************************************/
29
30 /* allocates a SystemMatrix of type type using the given matrix pattern
31 Values are initialized by zero.
32 if patternIsUnrolled and type & MATRIX_FORMAT_BLK1, it is assumed that the pattern is allready unrolled to match the requested block size
33 and offsets otherwise unrolling and offset adjustment will be performed.
34 */
35
36 Paso_SystemMatrix* Paso_SystemMatrix_alloc(Paso_SystemMatrixType type,Paso_SystemMatrixPattern *pattern, int row_block_size, int col_block_size,
37 const bool_t patternIsUnrolled) {
38
39 Paso_SystemMatrix*out=NULL;
40 dim_t n_norm,i;
41 Paso_SystemMatrixType pattern_format_out;
42 bool_t unroll=FALSE;
43
44 pattern_format_out= (type & MATRIX_FORMAT_OFFSET1)? MATRIX_FORMAT_OFFSET1: MATRIX_FORMAT_DEFAULT;
45 Esys_resetError();
46 if (patternIsUnrolled) {
47 if ( ! XNOR(type & MATRIX_FORMAT_OFFSET1, pattern->type & MATRIX_FORMAT_OFFSET1) ) {
48 Esys_setError(TYPE_ERROR,"Paso_SystemMatrix_alloc: requested offset and pattern offset does not match.");
49 return NULL;
50 }
51 }
52 /* do we need to apply unrolling ? */
53 unroll
54 /* we don't like non-square blocks */
55 = (row_block_size!=col_block_size)
56 /* or any block size bigger than 3 */
57 || (col_block_size>3)
58 /* or if lock size one requested and the block size is not 1 */
59 || ((type & MATRIX_FORMAT_BLK1) && (col_block_size>1) )
60 /* or the offsets are wrong */
61 || ((type & MATRIX_FORMAT_OFFSET1) != ( pattern->type & MATRIX_FORMAT_OFFSET1));
62
63 out=MEMALLOC(1,Paso_SystemMatrix);
64 if (! Esys_checkPtr(out)) {
65 out->type=type;
66 out->pattern=NULL;
67 out->row_distribution=NULL;
68 out->col_distribution=NULL;
69 out->mpi_info=Esys_MPIInfo_getReference(pattern->mpi_info);
70 out->row_coupler=NULL;
71 out->col_coupler=NULL;
72 out->mainBlock=NULL;
73 out->row_coupleBlock=NULL;
74 out->col_coupleBlock=NULL;
75 out->is_balanced=FALSE;
76 out->balance_vector=NULL;
77 out->solver_package=PASO_PASO;
78 out->solver_p=NULL;
79 out->trilinos_data=NULL;
80 out->reference_counter=1;
81 out->logical_row_block_size=row_block_size;
82 out->logical_col_block_size=col_block_size;
83
84
85 if (type & MATRIX_FORMAT_CSC) {
86 if (unroll) {
87 if (patternIsUnrolled) {
88 out->pattern=Paso_SystemMatrixPattern_getReference(pattern);
89 } else {
90 out->pattern=Paso_SystemMatrixPattern_unrollBlocks(pattern,pattern_format_out,col_block_size,row_block_size);
91 }
92 out->row_block_size=1;
93 out->col_block_size=1;
94 } else {
95 out->pattern=Paso_SystemMatrixPattern_unrollBlocks(pattern,pattern_format_out,1,1);
96 out->row_block_size=row_block_size;
97 out->col_block_size=col_block_size;
98 }
99 if (Esys_noError()) {
100 out->row_distribution=Paso_Distribution_getReference(out->pattern->input_distribution);
101 out->col_distribution=Paso_Distribution_getReference(out->pattern->output_distribution);
102 }
103 } else {
104 if (unroll) {
105 if (patternIsUnrolled) {
106 out->pattern=Paso_SystemMatrixPattern_getReference(pattern);
107 } else {
108 out->pattern=Paso_SystemMatrixPattern_unrollBlocks(pattern,pattern_format_out,row_block_size,col_block_size);
109 }
110 out->row_block_size=1;
111 out->col_block_size=1;
112 } else {
113 out->pattern=Paso_SystemMatrixPattern_unrollBlocks(pattern,pattern_format_out,1,1);
114 out->row_block_size=row_block_size;
115 out->col_block_size=col_block_size;
116 }
117 if (Esys_noError()) {
118 out->row_distribution=Paso_Distribution_getReference(out->pattern->output_distribution);
119 out->col_distribution=Paso_Distribution_getReference(out->pattern->input_distribution);
120 }
121 }
122 if (Esys_noError()) {
123 out->block_size=out->row_block_size*out->col_block_size;
124 out->col_coupler=Paso_Coupler_alloc(pattern->col_connector,out->col_block_size);
125 out->row_coupler=Paso_Coupler_alloc(pattern->row_connector,out->row_block_size);
126 /* this should be bypassed if trilinos is used */
127 if (type & MATRIX_FORMAT_TRILINOS_CRS) {
128 #ifdef TRILINOS
129 out->trilinos_data=Paso_TRILINOS_alloc();
130 #endif
131 } else {
132 out->solver_package=PASO_PASO;
133 out->mainBlock=Paso_SparseMatrix_alloc(type,out->pattern->mainPattern,row_block_size,col_block_size,TRUE);
134 out->col_coupleBlock=Paso_SparseMatrix_alloc(type,out->pattern->col_couplePattern,row_block_size,col_block_size,TRUE);
135 out->row_coupleBlock=Paso_SparseMatrix_alloc(type,out->pattern->row_couplePattern,row_block_size,col_block_size,TRUE);
136 if (Esys_noError()) {
137 /* allocate memory for matrix entries */
138 n_norm = MAX(out->mainBlock->numCols * out->col_block_size, out->mainBlock->numRows * out->row_block_size);
139 out->balance_vector=MEMALLOC(n_norm,double);
140 out->is_balanced=FALSE;
141 if (! Esys_checkPtr(out->balance_vector)) {
142 #pragma omp parallel for private(i) schedule(static)
143 for (i=0;i<n_norm;++i) out->balance_vector[i]=1.;
144 }
145 }
146 }
147 }
148 }
149 /* all done: */
150 if (! Esys_noError()) {
151 Paso_SystemMatrix_free(out);
152 return NULL;
153 } else {
154 return out;
155 }
156 }
157
158 /* returns a reference to Paso_SystemMatrix in */
159
160 Paso_SystemMatrix* Paso_SystemMatrix_getReference(Paso_SystemMatrix* in) {
161 if (in!=NULL) ++(in->reference_counter);
162 return in;
163 }
164
165 /* deallocates a SystemMatrix: */
166
167 void Paso_SystemMatrix_free(Paso_SystemMatrix* in) {
168 if (in!=NULL) {
169 in->reference_counter--;
170 if (in->reference_counter<=0) {
171 Paso_solve_free(in);
172 Paso_SystemMatrixPattern_free(in->pattern);
173 Paso_Distribution_free(in->row_distribution);
174 Paso_Distribution_free(in->col_distribution);
175 Esys_MPIInfo_free(in->mpi_info);
176 Paso_Coupler_free(in->row_coupler);
177 Paso_Coupler_free(in->col_coupler);
178 Paso_SparseMatrix_free(in->mainBlock);
179 Paso_SparseMatrix_free(in->col_coupleBlock);
180 Paso_SparseMatrix_free(in->row_coupleBlock);
181 MEMFREE(in->balance_vector);
182 Paso_solve_free(in);
183 #ifdef TRILINOS
184 Paso_TRILINOS_free(in->trilinos_data);
185 #endif
186 MEMFREE(in);
187 #ifdef Paso_TRACE
188 printf("Paso_SystemMatrix_free: system matrix as been deallocated.\n");
189 #endif
190 }
191 }
192 }
193 void Paso_SystemMatrix_startCollect(Paso_SystemMatrix* A,const double* in)
194 {
195 Paso_SystemMatrix_startColCollect(A,in);
196 }
197 double* Paso_SystemMatrix_finishCollect(Paso_SystemMatrix* A)
198 {
199 return Paso_SystemMatrix_finishColCollect(A);
200 }
201
202 void Paso_SystemMatrix_startColCollect(Paso_SystemMatrix* A,const double* in)
203 {
204 Paso_Coupler_startCollect(A->col_coupler, in);
205 }
206 double* Paso_SystemMatrix_finishColCollect(Paso_SystemMatrix* A)
207 {
208 Paso_Coupler_finishCollect(A->col_coupler);
209 return A->col_coupler->recv_buffer;
210 }
211 void Paso_SystemMatrix_startRowCollect(Paso_SystemMatrix* A,const double* in)
212 {
213 Paso_Coupler_startCollect(A->row_coupler, in);
214 }
215 double* Paso_SystemMatrix_finishRowCollect(Paso_SystemMatrix* A)
216 {
217 Paso_Coupler_finishCollect(A->row_coupler);
218 return A->row_coupler->recv_buffer;
219 }
220
221 dim_t Paso_SystemMatrix_getTotalNumRows(const Paso_SystemMatrix* A){
222 return A->mainBlock->numRows * A->row_block_size;
223 }
224
225 dim_t Paso_SystemMatrix_getTotalNumCols(const Paso_SystemMatrix* A){
226 return A->mainBlock->numCols * A->col_block_size;
227 }
228 dim_t Paso_SystemMatrix_getGlobalNumRows(Paso_SystemMatrix* A) {
229 if (A->type & MATRIX_FORMAT_CSC) {
230 return Paso_Distribution_getGlobalNumComponents(A->pattern->input_distribution);
231 } else {
232 return Paso_Distribution_getGlobalNumComponents(A->pattern->output_distribution);
233 }
234 }
235 dim_t Paso_SystemMatrix_getGlobalNumCols(Paso_SystemMatrix* A) {
236 if (A->type & MATRIX_FORMAT_CSC) {
237 return Paso_Distribution_getGlobalNumComponents(A->pattern->output_distribution);
238 } else {
239 return Paso_Distribution_getGlobalNumComponents(A->pattern->input_distribution);
240 }
241
242 }
243 dim_t Paso_SystemMatrix_getNumOutput(Paso_SystemMatrix* A) {
244 return Paso_SystemMatrixPattern_getNumOutput(A->pattern);
245 }
246
247 index_t* Paso_SystemMatrix_borrowMainDiagonalPointer(Paso_SystemMatrix * A_p)
248 {
249 index_t* out=NULL;
250 int fail=0;
251 out=Paso_SparseMatrix_borrowMainDiagonalPointer(A_p->mainBlock);
252 if (out==NULL) fail=1;
253 #ifdef ESYS_MPI
254 {
255 int fail_loc = fail;
256 MPI_Allreduce(&fail_loc, &fail, 1, MPI_INT, MPI_MAX, A_p->mpi_info->comm);
257 }
258 #endif
259 if (fail>0) Esys_setError(VALUE_ERROR, "Paso_SystemMatrix_borrowMainDiagonalPointer: no main diagonal");
260 return out;
261 }
262
263 void Paso_SystemMatrix_makeZeroRowSums(Paso_SystemMatrix * A_p, double* left_over)
264 {
265 index_t ir, ib, irow;
266 register double rtmp1, rtmp2;
267 const dim_t n = Paso_SystemMatrixPattern_getNumOutput(A_p->pattern);
268 const dim_t nblk = A_p->block_size;
269 const dim_t blk = A_p->row_block_size;
270 const index_t* main_ptr=Paso_SystemMatrix_borrowMainDiagonalPointer(A_p);
271
272
273 Paso_SystemMatrix_rowSum(A_p, left_over); /* left_over now hold the row sum */
274
275 #pragma omp parallel for private(ir,ib, rtmp1, rtmp2) schedule(static)
276 for (ir=0;ir< n;ir++) {
277 for (ib=0;ib<blk; ib++) {
278 irow=ib+blk*ir;
279 rtmp1=left_over[irow];
280 rtmp2=A_p->mainBlock->val[main_ptr[ir]*nblk+ib+blk*ib];
281 A_p->mainBlock->val[main_ptr[ir]*nblk+ib+blk*ib] = -rtmp1;
282 left_over[irow]=rtmp2+rtmp1;
283 }
284 }
285 }
286 void Paso_SystemMatrix_copyBlockFromMainDiagonal(Paso_SystemMatrix * A_p, double* out)
287 {
288 Paso_SparseMatrix_copyBlockFromMainDiagonal(A_p->mainBlock, out);
289 return;
290 }
291 void Paso_SystemMatrix_copyBlockToMainDiagonal(Paso_SystemMatrix * A_p, const double* in)
292 {
293 Paso_SparseMatrix_copyBlockToMainDiagonal(A_p->mainBlock, in);
294 return;
295 }
296 void Paso_SystemMatrix_copyFromMainDiagonal(Paso_SystemMatrix * A_p, double* out)
297 {
298 Paso_SparseMatrix_copyFromMainDiagonal(A_p->mainBlock, out);
299 return;
300 }
301 void Paso_SystemMatrix_copyToMainDiagonal(Paso_SystemMatrix * A_p, const double* in)
302 {
303 Paso_SparseMatrix_copyToMainDiagonal(A_p->mainBlock, in);
304 return;
305 }
306
307 void Paso_SystemMatrix_setPreconditioner(Paso_SystemMatrix* A,Paso_Options* options) {
308 if (A->solver_p==NULL) {
309 A->solver_p=Paso_Preconditioner_alloc(A,options);
310 }
311 }
312
313 /* applies the preconditioner */
314 /* has to be called within a parallel reqion */
315 /* barrier synchronization is performed before the evaluation to make sure that the input vector is available */
316 void Paso_SystemMatrix_solvePreconditioner(Paso_SystemMatrix* A,double* x,double* b){
317 Paso_Preconditioner* prec=(Paso_Preconditioner*) A->solver_p;
318 Paso_Preconditioner_solve(prec, A,x,b);
319 }
320 void Paso_SystemMatrix_freePreconditioner(Paso_SystemMatrix* A) {
321 Paso_Preconditioner* prec=NULL;
322 if (A!=NULL) {
323 prec=(Paso_Preconditioner*) A->solver_p;
324 Paso_Preconditioner_free(prec);
325 A->solver_p=NULL;
326 }
327 }

Properties

Name Value
svn:eol-style native
svn:keywords Author Date Id Revision

  ViewVC Help
Powered by ViewVC 1.1.26