/[escript]/trunk-mpi-branch/finley/src/Assemble_PDE_System2_2D.c
ViewVC logotype

Contents of /trunk-mpi-branch/finley/src/Assemble_PDE_System2_2D.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1253 - (show annotations)
Fri Aug 17 04:09:29 2007 UTC (12 years, 9 months ago) by gross
File MIME type: text/plain
File size: 17392 byte(s)
finally everthing for MPI is in the code. now testing is needed
1 /*
2 ************************************************************
3 * Copyright 2006 by ACcESS MNRF *
4 * *
5 * http://www.access.edu.au *
6 * Primary Business: Queensland, Australia *
7 * Licensed under the Open Software License version 3.0 *
8 * http://www.opensource.org/licenses/osl-3.0.php *
9 * *
10 ************************************************************
11 */
12
13 /**************************************************************/
14
15 /* assembles the system of numEq PDEs into the stiffness matrix S right hand side F */
16 /* the shape functions for test and solution must be identical */
17
18
19 /* -(A_{k,i,m,j} u_m,j)_i-(B_{k,i,m} u_m)_i+C_{k,m,j} u_m,j-D_{k,m} u_m and -(X_{k,i})_i + Y_k */
20
21 /* u has p.numComp components in a 2D domain. The shape functions for test and solution must be identical */
22 /* and row_NS == row_NN */
23
24 /* Shape of the coefficients: */
25
26 /* A = p.numEqu x 2 x p.numComp x 2 */
27 /* B = 2 x p.numEqu x p.numComp */
28 /* C = p.numEqu x 2 x p.numComp */
29 /* D = p.numEqu x p.numComp */
30 /* X = p.numEqu x 2 */
31 /* Y = p.numEqu */
32
33
34 /**************************************************************/
35
36 /* Author: gross@access.edu.au */
37 /* Version: $Id:$ */
38
39 /**************************************************************/
40
41
42 #include "Assemble.h"
43 #include "Util.h"
44 #ifdef _OPENMP
45 #include <omp.h>
46 #endif
47
48
49 /**************************************************************/
50
51 void Finley_Assemble_PDE_System2_2D(Assemble_Parameters p, Finley_ElementFile* elements,
52 Paso_SystemMatrix* Mat, escriptDataC* F,
53 escriptDataC* A, escriptDataC* B, escriptDataC* C, escriptDataC* D, escriptDataC* X, escriptDataC* Y) {
54
55 #define DIM 2
56 index_t color;
57 dim_t e;
58 double *EM_S, *EM_F, *Vol, *DSDX, *A_p, *B_p, *C_p, *D_p, *X_p, *Y_p;
59 index_t *row_index;
60 register dim_t q, s,r,k,m;
61 register double rtmp, rtmp0, rtmp1, rtmp00, rtmp10, rtmp01, rtmp11;
62 bool_t add_EM_F, add_EM_S;
63
64 bool_t extendedA=isExpanded(A);
65 bool_t extendedB=isExpanded(B);
66 bool_t extendedC=isExpanded(C);
67 bool_t extendedD=isExpanded(D);
68 bool_t extendedX=isExpanded(X);
69 bool_t extendedY=isExpanded(Y);
70 double *F_p=getSampleData(F,0);
71 double *S=p.row_jac->ReferenceElement->S;
72 dim_t len_EM_S=p.row_NN*p.col_NN*p.numEqu*p.numComp;
73 dim_t len_EM_F=p.row_NN*p.numEqu;
74
75
76 #pragma omp parallel private(color,EM_S, EM_F, Vol, DSDX, A_p, B_p, C_p, D_p, X_p, Y_p,row_index,q, s,r,k,m,rtmp, rtmp0, rtmp1, rtmp00, rtmp10, rtmp01, rtmp11,add_EM_F, add_EM_S)
77 {
78 EM_S=THREAD_MEMALLOC(len_EM_S,double);
79 EM_F=THREAD_MEMALLOC(len_EM_F,double);
80 row_index=THREAD_MEMALLOC(p.row_NN,index_t);
81
82 if (!Finley_checkPtr(EM_S) && !Finley_checkPtr(EM_F) && !Finley_checkPtr(row_index) ) {
83
84 for (color=elements->minColor;color<=elements->maxColor;color++) {
85 /* open loop over all elements: */
86 #pragma omp for private(e) schedule(static)
87 for(e=0;e<elements->numElements;e++){
88 if (elements->Color[e]==color) {
89 Vol=&(p.row_jac->volume[INDEX2(0,e,p.numQuad)]);
90 DSDX=&(p.row_jac->DSDX[INDEX4(0,0,0,e,p.row_NN,DIM,p.numQuad)]);
91 for (q=0;q<len_EM_S;++q) EM_S[q]=0;
92 for (q=0;q<len_EM_F;++q) EM_F[q]=0;
93 add_EM_F=FALSE;
94 add_EM_S=FALSE;
95 /**************************************************************/
96 /* process A: */
97 /**************************************************************/
98 A_p=getSampleData(A,e);
99 if (NULL!=A_p) {
100 add_EM_S=TRUE;
101 if (extendedA) {
102 for (s=0;s<p.row_NS;s++) {
103 for (r=0;r<p.col_NS;r++) {
104 for (k=0;k<p.numEqu;k++) {
105 for (m=0;m<p.numComp;m++) {
106 rtmp=0;
107 for (q=0;q<p.numQuad;q++) {
108 rtmp+=Vol[q]* (
109 DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*A_p[INDEX5(k,0,m,0,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
110 +DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*A_p[INDEX5(k,0,m,1,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]
111 +DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*A_p[INDEX5(k,1,m,0,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
112 +DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*A_p[INDEX5(k,1,m,1,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]);
113 }
114 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
115 }
116 }
117 }
118 }
119 } else {
120 for (s=0;s<p.row_NS;s++) {
121 for (r=0;r<p.col_NS;r++) {
122 rtmp00=0;
123 rtmp01=0;
124 rtmp10=0;
125 rtmp11=0;
126 for (q=0;q<p.numQuad;q++) {
127 rtmp0=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
128 rtmp1=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
129 rtmp00+=rtmp0*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
130 rtmp01+=rtmp0*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
131 rtmp10+=rtmp1*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
132 rtmp11+=rtmp1*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
133 }
134 for (k=0;k<p.numEqu;k++) {
135 for (m=0;m<p.numComp;m++) {
136 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=
137 rtmp00*A_p[INDEX4(k,0,m,0,p.numEqu,DIM,p.numComp)]
138 +rtmp01*A_p[INDEX4(k,0,m,1,p.numEqu,DIM,p.numComp)]
139 +rtmp10*A_p[INDEX4(k,1,m,0,p.numEqu,DIM,p.numComp)]
140 +rtmp11*A_p[INDEX4(k,1,m,1,p.numEqu,DIM,p.numComp)];
141 }
142 }
143 }
144 }
145 }
146 }
147 /**************************************************************/
148 /* process B: */
149 /**************************************************************/
150 B_p=getSampleData(B,e);
151 if (NULL!=B_p) {
152 add_EM_S=TRUE;
153 if (extendedB) {
154 for (s=0;s<p.row_NS;s++) {
155 for (r=0;r<p.col_NS;r++) {
156 for (k=0;k<p.numEqu;k++) {
157 for (m=0;m<p.numComp;m++) {
158 rtmp=0;
159 for (q=0;q<p.numQuad;q++) {
160 rtmp+=Vol[q]*S[INDEX2(r,q,p.row_NS)]*
161 ( DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*B_p[INDEX4(k,0,m,q,p.numEqu,DIM,p.numComp)]
162 + DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*B_p[INDEX4(k,1,m,q,p.numEqu,DIM,p.numComp)]);
163 }
164 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
165 }
166 }
167 }
168 }
169 } else {
170 for (s=0;s<p.row_NS;s++) {
171 for (r=0;r<p.col_NS;r++) {
172 rtmp0=0;
173 rtmp1=0;
174 for (q=0;q<p.numQuad;q++) {
175 rtmp=Vol[q]*S[INDEX2(r,q,p.row_NS)];
176 rtmp0+=rtmp*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
177 rtmp1+=rtmp*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
178 }
179 for (k=0;k<p.numEqu;k++) {
180 for (m=0;m<p.numComp;m++) {
181 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+= rtmp0*B_p[INDEX3(k,0,m,p.numEqu,DIM)]
182 + rtmp1*B_p[INDEX3(k,1,m,p.numEqu,DIM)];
183 }
184 }
185 }
186 }
187 }
188 }
189 /**************************************************************/
190 /* process C: */
191 /**************************************************************/
192 C_p=getSampleData(C,e);
193 if (NULL!=C_p) {
194 add_EM_S=TRUE;
195 if (extendedC) {
196 for (s=0;s<p.row_NS;s++) {
197 for (r=0;r<p.col_NS;r++) {
198 for (k=0;k<p.numEqu;k++) {
199 for (m=0;m<p.numComp;m++) {
200 rtmp=0;
201 for (q=0;q<p.numQuad;q++) {
202 rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*
203 ( C_p[INDEX4(k,m,0,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
204 + C_p[INDEX4(k,m,1,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]);
205 }
206 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
207 }
208 }
209 }
210 }
211 } else {
212 for (s=0;s<p.row_NS;s++) {
213 for (r=0;r<p.col_NS;r++) {
214 rtmp0=0;
215 rtmp1=0;
216 for (q=0;q<p.numQuad;q++) {
217 rtmp=Vol[q]*S[INDEX2(s,q,p.row_NS)];
218 rtmp0+=rtmp*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
219 rtmp1+=rtmp*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
220 }
221 for (k=0;k<p.numEqu;k++) {
222 for (m=0;m<p.numComp;m++) {
223 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp0*C_p[INDEX3(k,m,0,p.numEqu,p.numComp)]
224 +rtmp1*C_p[INDEX3(k,m,1,p.numEqu,p.numComp)];
225 }
226 }
227 }
228 }
229 }
230 }
231 /************************************************************* */
232 /* process D */
233 /**************************************************************/
234 D_p=getSampleData(D,e);
235 if (NULL!=D_p) {
236 add_EM_S=TRUE;
237 if (extendedD) {
238 for (s=0;s<p.row_NS;s++) {
239 for (r=0;r<p.col_NS;r++) {
240 for (k=0;k<p.numEqu;k++) {
241 for (m=0;m<p.numComp;m++) {
242 rtmp=0;
243 for (q=0;q<p.numQuad;q++) {
244 rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*D_p[INDEX3(k,m,q,p.numEqu,p.numComp)]*S[INDEX2(r,q,p.row_NS)];
245 }
246 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
247 }
248 }
249 }
250 }
251 } else {
252 for (s=0;s<p.row_NS;s++) {
253 for (r=0;r<p.col_NS;r++) {
254 rtmp=0;
255 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*S[INDEX2(r,q,p.row_NS)];
256 for (k=0;k<p.numEqu;k++) {
257 for (m=0;m<p.numComp;m++) {
258 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp*D_p[INDEX2(k,m,p.numEqu)];
259 }
260 }
261 }
262 }
263 }
264 }
265 /**************************************************************/
266 /* process X: */
267 /**************************************************************/
268 X_p=getSampleData(X,e);
269 if (NULL!=X_p) {
270 add_EM_F=TRUE;
271 if (extendedX) {
272 for (s=0;s<p.row_NS;s++) {
273 for (k=0;k<p.numEqu;k++) {
274 rtmp=0;
275 for (q=0;q<p.numQuad;q++) {
276 rtmp+=Vol[q]*(DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*X_p[INDEX3(k,0,q,p.numEqu,DIM)]
277 +DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*X_p[INDEX3(k,1,q,p.numEqu,DIM)]);
278 }
279 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
280 }
281 }
282 } else {
283 for (s=0;s<p.row_NS;s++) {
284 rtmp0=0;
285 rtmp1=0;
286 for (q=0;q<p.numQuad;q++) {
287 rtmp0+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
288 rtmp1+=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
289 }
290 for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp0*X_p[INDEX2(k,0,p.numEqu)]+rtmp1*X_p[INDEX2(k,1,p.numEqu)];
291 }
292 }
293 }
294 /**************************************************************/
295 /* process Y: */
296 /**************************************************************/
297 Y_p=getSampleData(Y,e);
298 if (NULL!=Y_p) {
299 add_EM_F=TRUE;
300 if (extendedY) {
301 for (s=0;s<p.row_NS;s++) {
302 for (k=0;k<p.numEqu;k++) {
303 rtmp=0;
304 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*Y_p[INDEX2(k,q,p.numEqu)];
305 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
306 }
307 }
308 } else {
309 for (s=0;s<p.row_NS;s++) {
310 rtmp=0;
311 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)];
312 for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp*Y_p[k];
313 }
314 }
315 }
316 /***********************************************************************************************/
317 /* add the element matrices onto the matrix and right hand side */
318 /***********************************************************************************************/
319 for (q=0;q<p.row_NN;q++) row_index[q]=p.row_DOF[elements->Nodes[INDEX2(p.row_node[q],e,p.NN)]];
320 if (add_EM_F) Finley_Util_AddScatter(p.row_NN,row_index,p.numEqu,EM_F,F_p, p.row_DOF_UpperBound);
321 if (add_EM_S) Finley_Assemble_addToSystemMatrix(Mat,p.row_NN,row_index,p.numEqu,p.col_NN,row_index,p.numComp,EM_S);
322
323 } /* end color check */
324 } /* end element loop */
325 } /* end color loop */
326
327 THREAD_MEMFREE(EM_S);
328 THREAD_MEMFREE(EM_F);
329 THREAD_MEMFREE(row_index);
330
331 } /* end of pointer check */
332 } /* end parallel region */
333 }
334 /*
335 * $Log$
336 */

  ViewVC Help
Powered by ViewVC 1.1.26