/[escript]/branches/arrexp_2137_win_merge/finley/src/Assemble_PDE_System2_3D.c
ViewVC logotype

Contents of /branches/arrexp_2137_win_merge/finley/src/Assemble_PDE_System2_3D.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 2222 - (show annotations)
Tue Jan 20 04:52:39 2009 UTC (11 years, 1 month ago) by jfenwick
File MIME type: text/plain
File size: 21403 byte(s)
Saving work
1
2 /*******************************************************
3 *
4 * Copyright (c) 2003-2008 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 /* assembles the system of numEq PDEs into the stiffness matrix S right hand side F */
18 /* the shape functions for test and solution must be identical */
19
20
21 /* -(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 */
22
23 /* u has p.numComp components in a 3D domain. The shape functions for test and solution must be identical */
24 /* and row_NS == row_NN */
25
26 /* Shape of the coefficients: */
27
28 /* A = p.numEqu x 3 x p.numComp x 3 */
29 /* B = 3 x p.numEqu x p.numComp */
30 /* C = p.numEqu x 3 x p.numComp */
31 /* D = p.numEqu x p.numComp */
32 /* X = p.numEqu x 3 */
33 /* Y = p.numEqu */
34
35
36 /**************************************************************/
37
38
39 #include "Assemble.h"
40 #include "Util.h"
41 #ifdef _OPENMP
42 #include <omp.h>
43 #endif
44
45
46 /**************************************************************/
47
48 void Finley_Assemble_PDE_System2_3D(Assemble_Parameters p, Finley_ElementFile* elements,
49 Paso_SystemMatrix* Mat, escriptDataC* F,
50 escriptDataC* A, escriptDataC* B, escriptDataC* C, escriptDataC* D, escriptDataC* X, escriptDataC* Y) {
51
52 #define DIM 3
53 index_t color;
54 dim_t e;
55 __const double *A_p, *B_p, *C_p, *D_p, *X_p, *Y_p;
56 double *EM_S, *EM_F, *Vol, *DSDX;
57 index_t *row_index;
58 register dim_t q, s,r,k,m;
59 register double rtmp, rtmp0, rtmp1, rtmp2, rtmp00, rtmp01, rtmp02, rtmp10, rtmp11, rtmp12, rtmp20, rtmp21, rtmp22;
60 bool_t add_EM_F, add_EM_S;
61
62 bool_t extendedA=isExpanded(A);
63 bool_t extendedB=isExpanded(B);
64 bool_t extendedC=isExpanded(C);
65 bool_t extendedD=isExpanded(D);
66 bool_t extendedX=isExpanded(X);
67 bool_t extendedY=isExpanded(Y);
68 double *F_p=(requireWrite(F), getSampleDataRW(F,0)); /* use comma, to get around the mixed code and declarations thing */
69 double *S=p.row_jac->ReferenceElement->S;
70 dim_t len_EM_S=p.row_NN*p.col_NN*p.numEqu*p.numComp;
71 dim_t len_EM_F=p.row_NN*p.numEqu;
72
73
74 #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, rtmp2, rtmp00, rtmp01, rtmp02, rtmp10, rtmp11, rtmp12, rtmp20, rtmp21, rtmp22,add_EM_F, add_EM_S)
75 {
76 void* ABuff=allocSampleBuffer(A);
77 void* BBuff=allocSampleBuffer(B);
78 void* CBuff=allocSampleBuffer(C);
79 void* DBuff=allocSampleBuffer(D);
80 void* XBuff=allocSampleBuffer(X);
81 void* YBuff=allocSampleBuffer(Y);
82 EM_S=THREAD_MEMALLOC(len_EM_S,double);
83 EM_F=THREAD_MEMALLOC(len_EM_F,double);
84 row_index=THREAD_MEMALLOC(p.row_NN,index_t);
85
86 if (!Finley_checkPtr(EM_S) && !Finley_checkPtr(EM_F) && !Finley_checkPtr(row_index) ) {
87
88 for (color=elements->minColor;color<=elements->maxColor;color++) {
89 /* open loop over all elements: */
90 #pragma omp for private(e) schedule(static)
91 for(e=0;e<elements->numElements;e++){
92 if (elements->Color[e]==color) {
93 Vol=&(p.row_jac->volume[INDEX2(0,e,p.numQuad)]);
94 DSDX=&(p.row_jac->DSDX[INDEX4(0,0,0,e,p.row_NN,DIM,p.numQuad)]);
95 for (q=0;q<len_EM_S;++q) EM_S[q]=0;
96 for (q=0;q<len_EM_F;++q) EM_F[q]=0;
97 add_EM_F=FALSE;
98 add_EM_S=FALSE;
99 /**************************************************************/
100 /* process A: */
101 /**************************************************************/
102 A_p=getSampleDataRO(A,e,ABuff);
103 if (NULL!=A_p) {
104 add_EM_S=TRUE;
105 if (extendedA) {
106 for (s=0;s<p.row_NS;s++) {
107 for (r=0;r<p.col_NS;r++) {
108 for (k=0;k<p.numEqu;k++) {
109 for (m=0;m<p.numComp;m++) {
110 rtmp=0;
111 for (q=0;q<p.numQuad;q++) {
112 rtmp+=Vol[q]*(
113 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)]
114 +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)]
115 +DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*A_p[INDEX5(k,0,m,2,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,2,q,p.row_NN,DIM)]
116 +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)]
117 +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)]
118 +DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*A_p[INDEX5(k,1,m,2,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,2,q,p.row_NN,DIM)]
119 +DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*A_p[INDEX5(k,2,m,0,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
120 +DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*A_p[INDEX5(k,2,m,1,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]
121 +DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*A_p[INDEX5(k,2,m,2,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,2,q,p.row_NN,DIM)]);
122
123 }
124 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
125 }
126 }
127 }
128 }
129 } else {
130 for (s=0;s<p.row_NS;s++) {
131 for (r=0;r<p.col_NS;r++) {
132 rtmp00=0;
133 rtmp01=0;
134 rtmp02=0;
135 rtmp10=0;
136 rtmp11=0;
137 rtmp12=0;
138 rtmp20=0;
139 rtmp21=0;
140 rtmp22=0;
141 for (q=0;q<p.numQuad;q++) {
142 rtmp0=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
143 rtmp00+=rtmp0*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
144 rtmp01+=rtmp0*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
145 rtmp02+=rtmp0*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
146
147 rtmp1=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
148 rtmp10+=rtmp1*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
149 rtmp11+=rtmp1*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
150 rtmp12+=rtmp1*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
151
152 rtmp2=Vol[q]*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
153 rtmp20+=rtmp2*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
154 rtmp21+=rtmp2*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
155 rtmp22+=rtmp2*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
156 }
157 for (k=0;k<p.numEqu;k++) {
158 for (m=0;m<p.numComp;m++) {
159 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+= rtmp00*A_p[INDEX4(k,0,m,0,p.numEqu,DIM,p.numComp)]
160 +rtmp01*A_p[INDEX4(k,0,m,1,p.numEqu,DIM,p.numComp)]
161 +rtmp02*A_p[INDEX4(k,0,m,2,p.numEqu,DIM,p.numComp)]
162 +rtmp10*A_p[INDEX4(k,1,m,0,p.numEqu,DIM,p.numComp)]
163 +rtmp11*A_p[INDEX4(k,1,m,1,p.numEqu,DIM,p.numComp)]
164 +rtmp12*A_p[INDEX4(k,1,m,2,p.numEqu,DIM,p.numComp)]
165 +rtmp20*A_p[INDEX4(k,2,m,0,p.numEqu,DIM,p.numComp)]
166 +rtmp21*A_p[INDEX4(k,2,m,1,p.numEqu,DIM,p.numComp)]
167 +rtmp22*A_p[INDEX4(k,2,m,2,p.numEqu,DIM,p.numComp)];
168 }
169 }
170 }
171 }
172 }
173 }
174 /**************************************************************/
175 /* process B: */
176 /**************************************************************/
177 B_p=getSampleDataRO(B,e,BBuff);
178 if (NULL!=B_p) {
179 add_EM_S=TRUE;
180 if (extendedB) {
181 for (s=0;s<p.row_NS;s++) {
182 for (r=0;r<p.col_NS;r++) {
183 for (k=0;k<p.numEqu;k++) {
184 for (m=0;m<p.numComp;m++) {
185 rtmp=0;
186 for (q=0;q<p.numQuad;q++) {
187 rtmp+=Vol[q]*S[INDEX2(r,q,p.row_NS)]*(
188 DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*B_p[INDEX4(k,0,m,q,p.numEqu,DIM,p.numComp)]
189 + DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*B_p[INDEX4(k,1,m,q,p.numEqu,DIM,p.numComp)]
190 + DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*B_p[INDEX4(k,2,m,q,p.numEqu,DIM,p.numComp)] );
191 }
192 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
193 }
194 }
195 }
196 }
197 } else {
198 for (s=0;s<p.row_NS;s++) {
199 for (r=0;r<p.col_NS;r++) {
200 rtmp0=0;
201 rtmp1=0;
202 rtmp2=0;
203 for (q=0;q<p.numQuad;q++) {
204 rtmp=Vol[q]*S[INDEX2(r,q,p.row_NS)];
205 rtmp0+=rtmp*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
206 rtmp1+=rtmp*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
207 rtmp2+=rtmp*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
208 }
209 for (k=0;k<p.numEqu;k++) {
210 for (m=0;m<p.numComp;m++) {
211 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp0*B_p[INDEX3(k,0,m,p.numEqu,DIM)]
212 +rtmp1*B_p[INDEX3(k,1,m,p.numEqu,DIM)]
213 +rtmp2*B_p[INDEX3(k,2,m,p.numEqu,DIM)];
214 }
215 }
216 }
217 }
218 }
219 }
220 /**************************************************************/
221 /* process C: */
222 /**************************************************************/
223 C_p=getSampleDataRO(C,e,CBuff);
224 if (NULL!=C_p) {
225 add_EM_S=TRUE;
226 if (extendedC) {
227 for (s=0;s<p.row_NS;s++) {
228 for (r=0;r<p.col_NS;r++) {
229 for (k=0;k<p.numEqu;k++) {
230 for (m=0;m<p.numComp;m++) {
231 rtmp=0;
232 for (q=0;q<p.numQuad;q++) {
233 rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*(
234 C_p[INDEX4(k,m,0,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
235 +C_p[INDEX4(k,m,1,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]
236 +C_p[INDEX4(k,m,2,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,2,q,p.row_NN,DIM)]);
237 }
238 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
239 }
240 }
241 }
242 }
243 } else {
244 for (s=0;s<p.row_NS;s++) {
245 for (r=0;r<p.col_NS;r++) {
246 rtmp0=0;
247 rtmp1=0;
248 rtmp2=0;
249 for (q=0;q<p.numQuad;q++) {
250 rtmp=Vol[q]*S[INDEX2(s,q,p.row_NS)];
251 rtmp0+=rtmp*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
252 rtmp1+=rtmp*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
253 rtmp2+=rtmp*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
254 }
255 for (k=0;k<p.numEqu;k++) {
256 for (m=0;m<p.numComp;m++) {
257 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)]
258 +rtmp1*C_p[INDEX3(k,m,1,p.numEqu,p.numComp)]
259 +rtmp2*C_p[INDEX3(k,m,2,p.numEqu,p.numComp)];
260 }
261 }
262 }
263 }
264 }
265 }
266 /************************************************************* */
267 /* process D */
268 /**************************************************************/
269 D_p=getSampleDataRO(D,e,DBuff);
270 if (NULL!=D_p) {
271 add_EM_S=TRUE;
272 if (extendedD) {
273 for (s=0;s<p.row_NS;s++) {
274 for (r=0;r<p.col_NS;r++) {
275 for (k=0;k<p.numEqu;k++) {
276 for (m=0;m<p.numComp;m++) {
277 rtmp=0;
278 for (q=0;q<p.numQuad;q++) {
279 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)];
280 }
281 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
282 }
283 }
284 }
285 }
286 } else {
287 for (s=0;s<p.row_NS;s++) {
288 for (r=0;r<p.col_NS;r++) {
289 rtmp=0;
290 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*S[INDEX2(r,q,p.row_NS)];
291 for (k=0;k<p.numEqu;k++) {
292 for (m=0;m<p.numComp;m++) {
293 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp*D_p[INDEX2(k,m,p.numEqu)];
294 }
295 }
296 }
297 }
298 }
299 }
300 /**************************************************************/
301 /* process X: */
302 /**************************************************************/
303 X_p=getSampleDataRO(X,e,XBuff);
304 if (NULL!=X_p) {
305 add_EM_F=TRUE;
306 if (extendedX) {
307 for (s=0;s<p.row_NS;s++) {
308 for (k=0;k<p.numEqu;k++) {
309 rtmp=0;
310 for (q=0;q<p.numQuad;q++) {
311 rtmp+=Vol[q]* ( DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*X_p[INDEX3(k,0,q,p.numEqu,DIM)]
312 + DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*X_p[INDEX3(k,1,q,p.numEqu,DIM)]
313 + DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*X_p[INDEX3(k,2,q,p.numEqu,DIM)]);
314 }
315 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
316 }
317 }
318 } else {
319 for (s=0;s<p.row_NS;s++) {
320 rtmp0=0;
321 rtmp1=0;
322 rtmp2=0;
323 for (q=0;q<p.numQuad;q++) {
324 rtmp0+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
325 rtmp1+=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
326 rtmp2+=Vol[q]*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
327 }
328 for (k=0;k<p.numEqu;k++) {
329 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp0*X_p[INDEX2(k,0,p.numEqu)]
330 +rtmp1*X_p[INDEX2(k,1,p.numEqu)]
331 +rtmp2*X_p[INDEX2(k,2,p.numEqu)];
332 }
333 }
334 }
335 }
336 /**************************************************************/
337 /* process Y: */
338 /**************************************************************/
339 Y_p=getSampleDataRO(Y,e,YBuff);
340 if (NULL!=Y_p) {
341 add_EM_F=TRUE;
342 if (extendedY) {
343 for (s=0;s<p.row_NS;s++) {
344 for (k=0;k<p.numEqu;k++) {
345 rtmp=0.;
346 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*Y_p[INDEX2(k,q,p.numEqu)];
347 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
348 }
349 }
350 } else {
351 for (s=0;s<p.row_NS;s++) {
352 rtmp=0;
353 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)];
354 for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp*Y_p[k];
355 }
356 }
357 }
358 /***********************************************************************************************/
359 /* add the element matrices onto the matrix and right hand side */
360 /***********************************************************************************************/
361 for (q=0;q<p.row_NN;q++) row_index[q]=p.row_DOF[elements->Nodes[INDEX2(p.row_node[q],e,p.NN)]];
362 if (add_EM_F) Finley_Util_AddScatter(p.row_NN,row_index,p.numEqu,EM_F,F_p, p.row_DOF_UpperBound);
363 if (add_EM_S) Finley_Assemble_addToSystemMatrix(Mat,p.row_NN,row_index,p.numEqu,p.col_NN,row_index,p.numComp,EM_S);
364
365 } /* end color check */
366 } /* end element loop */
367 } /* end color loop */
368
369 THREAD_MEMFREE(EM_S);
370 THREAD_MEMFREE(EM_F);
371 THREAD_MEMFREE(row_index);
372
373 } /* end of pointer check */
374 freeSampleBuffer(ABuff);
375 freeSampleBuffer(BBuff);
376 freeSampleBuffer(CBuff);
377 freeSampleBuffer(DBuff);
378 freeSampleBuffer(XBuff);
379 freeSampleBuffer(YBuff);
380 } /* end parallel region */
381 }
382 /*
383 * $Log$
384 */

  ViewVC Help
Powered by ViewVC 1.1.26