/[escript]/trunk/finley/src/Assemble_PDE_System2_3D.c
ViewVC logotype

Annotation of /trunk/finley/src/Assemble_PDE_System2_3D.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 2548 - (hide annotations)
Mon Jul 20 06:20:06 2009 UTC (9 years, 11 months ago) by jfenwick
File MIME type: text/plain
File size: 21366 byte(s)
Updating copyright notices
1 gross 798
2 ksteube 1312 /*******************************************************
3 ksteube 1811 *
4 jfenwick 2548 * Copyright (c) 2003-2009 by University of Queensland
5 ksteube 1811 * 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 ksteube 1312
14 ksteube 1811
15 gross 798 /**************************************************************/
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 gross 853 #ifdef _OPENMP
42     #include <omp.h>
43     #endif
44 gross 798
45 gross 853
46 gross 798 /**************************************************************/
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 jfenwick 2271 __const double *A_p, *B_p, *C_p, *D_p, *X_p, *Y_p;
56     double *EM_S, *EM_F, *Vol, *DSDX;
57 gross 853 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 gross 798 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 jfenwick 2271 double *F_p=(requireWrite(F), getSampleDataRW(F,0)); /* use comma, to get around the mixed code and declarations thing */
69 gross 798 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 jfenwick 2271 void* ABuff=allocSampleBuffer(A);
74     void* BBuff=allocSampleBuffer(B);
75     void* CBuff=allocSampleBuffer(C);
76     void* DBuff=allocSampleBuffer(D);
77     void* XBuff=allocSampleBuffer(X);
78     void* YBuff=allocSampleBuffer(Y);
79 gross 853 #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)
80 gross 798 {
81 gross 853 EM_S=THREAD_MEMALLOC(len_EM_S,double);
82     EM_F=THREAD_MEMALLOC(len_EM_F,double);
83     row_index=THREAD_MEMALLOC(p.row_NN,index_t);
84    
85     if (!Finley_checkPtr(EM_S) && !Finley_checkPtr(EM_F) && !Finley_checkPtr(row_index) ) {
86 gross 798
87 gross 853 for (color=elements->minColor;color<=elements->maxColor;color++) {
88     /* open loop over all elements: */
89     #pragma omp for private(e) schedule(static)
90     for(e=0;e<elements->numElements;e++){
91     if (elements->Color[e]==color) {
92     Vol=&(p.row_jac->volume[INDEX2(0,e,p.numQuad)]);
93     DSDX=&(p.row_jac->DSDX[INDEX4(0,0,0,e,p.row_NN,DIM,p.numQuad)]);
94     for (q=0;q<len_EM_S;++q) EM_S[q]=0;
95     for (q=0;q<len_EM_F;++q) EM_F[q]=0;
96     add_EM_F=FALSE;
97     add_EM_S=FALSE;
98     /**************************************************************/
99     /* process A: */
100     /**************************************************************/
101 jfenwick 2271 A_p=getSampleDataRO(A,e,ABuff);
102 gross 853 if (NULL!=A_p) {
103     add_EM_S=TRUE;
104     if (extendedA) {
105     for (s=0;s<p.row_NS;s++) {
106     for (r=0;r<p.col_NS;r++) {
107     for (k=0;k<p.numEqu;k++) {
108 gross 798 for (m=0;m<p.numComp;m++) {
109 gross 853 rtmp=0;
110     for (q=0;q<p.numQuad;q++) {
111     rtmp+=Vol[q]*(
112     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)]
113     +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)]
114     +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)]
115     +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)]
116     +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)]
117     +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)]
118     +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)]
119     +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)]
120     +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)]);
121    
122     }
123     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
124 gross 798 }
125 gross 853 }
126     }
127     }
128     } else {
129     for (s=0;s<p.row_NS;s++) {
130     for (r=0;r<p.col_NS;r++) {
131     rtmp00=0;
132     rtmp01=0;
133     rtmp02=0;
134     rtmp10=0;
135     rtmp11=0;
136     rtmp12=0;
137     rtmp20=0;
138     rtmp21=0;
139     rtmp22=0;
140 gross 798 for (q=0;q<p.numQuad;q++) {
141 gross 853 rtmp0=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
142     rtmp00+=rtmp0*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
143     rtmp01+=rtmp0*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
144     rtmp02+=rtmp0*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
145    
146     rtmp1=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
147     rtmp10+=rtmp1*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
148     rtmp11+=rtmp1*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
149     rtmp12+=rtmp1*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
150    
151     rtmp2=Vol[q]*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
152     rtmp20+=rtmp2*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
153     rtmp21+=rtmp2*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
154     rtmp22+=rtmp2*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
155 gross 798 }
156     for (k=0;k<p.numEqu;k++) {
157     for (m=0;m<p.numComp;m++) {
158 gross 853 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)]
159     +rtmp01*A_p[INDEX4(k,0,m,1,p.numEqu,DIM,p.numComp)]
160     +rtmp02*A_p[INDEX4(k,0,m,2,p.numEqu,DIM,p.numComp)]
161     +rtmp10*A_p[INDEX4(k,1,m,0,p.numEqu,DIM,p.numComp)]
162     +rtmp11*A_p[INDEX4(k,1,m,1,p.numEqu,DIM,p.numComp)]
163     +rtmp12*A_p[INDEX4(k,1,m,2,p.numEqu,DIM,p.numComp)]
164     +rtmp20*A_p[INDEX4(k,2,m,0,p.numEqu,DIM,p.numComp)]
165     +rtmp21*A_p[INDEX4(k,2,m,1,p.numEqu,DIM,p.numComp)]
166     +rtmp22*A_p[INDEX4(k,2,m,2,p.numEqu,DIM,p.numComp)];
167 gross 798 }
168     }
169 gross 853 }
170     }
171     }
172     }
173     /**************************************************************/
174     /* process B: */
175     /**************************************************************/
176 jfenwick 2271 B_p=getSampleDataRO(B,e,BBuff);
177 gross 853 if (NULL!=B_p) {
178     add_EM_S=TRUE;
179     if (extendedB) {
180     for (s=0;s<p.row_NS;s++) {
181     for (r=0;r<p.col_NS;r++) {
182     for (k=0;k<p.numEqu;k++) {
183     for (m=0;m<p.numComp;m++) {
184     rtmp=0;
185     for (q=0;q<p.numQuad;q++) {
186     rtmp+=Vol[q]*S[INDEX2(r,q,p.row_NS)]*(
187     DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*B_p[INDEX4(k,0,m,q,p.numEqu,DIM,p.numComp)]
188     + DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*B_p[INDEX4(k,1,m,q,p.numEqu,DIM,p.numComp)]
189     + DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*B_p[INDEX4(k,2,m,q,p.numEqu,DIM,p.numComp)] );
190     }
191     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
192     }
193     }
194     }
195     }
196     } else {
197     for (s=0;s<p.row_NS;s++) {
198     for (r=0;r<p.col_NS;r++) {
199     rtmp0=0;
200     rtmp1=0;
201     rtmp2=0;
202     for (q=0;q<p.numQuad;q++) {
203     rtmp=Vol[q]*S[INDEX2(r,q,p.row_NS)];
204     rtmp0+=rtmp*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
205     rtmp1+=rtmp*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
206     rtmp2+=rtmp*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
207     }
208     for (k=0;k<p.numEqu;k++) {
209     for (m=0;m<p.numComp;m++) {
210     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp0*B_p[INDEX3(k,0,m,p.numEqu,DIM)]
211     +rtmp1*B_p[INDEX3(k,1,m,p.numEqu,DIM)]
212     +rtmp2*B_p[INDEX3(k,2,m,p.numEqu,DIM)];
213     }
214     }
215     }
216     }
217 gross 798 }
218     }
219 gross 853 /**************************************************************/
220     /* process C: */
221     /**************************************************************/
222 jfenwick 2271 C_p=getSampleDataRO(C,e,CBuff);
223 gross 853 if (NULL!=C_p) {
224     add_EM_S=TRUE;
225     if (extendedC) {
226     for (s=0;s<p.row_NS;s++) {
227     for (r=0;r<p.col_NS;r++) {
228     for (k=0;k<p.numEqu;k++) {
229     for (m=0;m<p.numComp;m++) {
230     rtmp=0;
231     for (q=0;q<p.numQuad;q++) {
232     rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*(
233     C_p[INDEX4(k,m,0,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_NN,DIM)]
234     +C_p[INDEX4(k,m,1,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,1,q,p.row_NN,DIM)]
235     +C_p[INDEX4(k,m,2,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,2,q,p.row_NN,DIM)]);
236     }
237     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
238 gross 798 }
239 gross 853 }
240     }
241     }
242     } else {
243     for (s=0;s<p.row_NS;s++) {
244     for (r=0;r<p.col_NS;r++) {
245     rtmp0=0;
246     rtmp1=0;
247     rtmp2=0;
248     for (q=0;q<p.numQuad;q++) {
249     rtmp=Vol[q]*S[INDEX2(s,q,p.row_NS)];
250     rtmp0+=rtmp*DSDX[INDEX3(r,0,q,p.row_NN,DIM)];
251     rtmp1+=rtmp*DSDX[INDEX3(r,1,q,p.row_NN,DIM)];
252     rtmp2+=rtmp*DSDX[INDEX3(r,2,q,p.row_NN,DIM)];
253     }
254     for (k=0;k<p.numEqu;k++) {
255     for (m=0;m<p.numComp;m++) {
256     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)]
257     +rtmp1*C_p[INDEX3(k,m,1,p.numEqu,p.numComp)]
258     +rtmp2*C_p[INDEX3(k,m,2,p.numEqu,p.numComp)];
259     }
260     }
261     }
262     }
263     }
264     }
265     /************************************************************* */
266     /* process D */
267     /**************************************************************/
268 jfenwick 2271 D_p=getSampleDataRO(D,e,DBuff);
269 gross 853 if (NULL!=D_p) {
270     add_EM_S=TRUE;
271     if (extendedD) {
272     for (s=0;s<p.row_NS;s++) {
273     for (r=0;r<p.col_NS;r++) {
274     for (k=0;k<p.numEqu;k++) {
275     for (m=0;m<p.numComp;m++) {
276     rtmp=0;
277     for (q=0;q<p.numQuad;q++) {
278     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)];
279     }
280     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp;
281     }
282     }
283     }
284     }
285     } else {
286     for (s=0;s<p.row_NS;s++) {
287     for (r=0;r<p.col_NS;r++) {
288     rtmp=0;
289     for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*S[INDEX2(r,q,p.row_NS)];
290 gross 798 for (k=0;k<p.numEqu;k++) {
291 gross 853 for (m=0;m<p.numComp;m++) {
292     EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_NN)]+=rtmp*D_p[INDEX2(k,m,p.numEqu)];
293     }
294 gross 798 }
295 gross 853 }
296     }
297     }
298     }
299     /**************************************************************/
300     /* process X: */
301     /**************************************************************/
302 jfenwick 2271 X_p=getSampleDataRO(X,e,XBuff);
303 gross 853 if (NULL!=X_p) {
304     add_EM_F=TRUE;
305     if (extendedX) {
306     for (s=0;s<p.row_NS;s++) {
307     for (k=0;k<p.numEqu;k++) {
308 gross 798 rtmp=0;
309     for (q=0;q<p.numQuad;q++) {
310 gross 853 rtmp+=Vol[q]* ( DSDX[INDEX3(s,0,q,p.row_NN,DIM)]*X_p[INDEX3(k,0,q,p.numEqu,DIM)]
311     + DSDX[INDEX3(s,1,q,p.row_NN,DIM)]*X_p[INDEX3(k,1,q,p.numEqu,DIM)]
312     + DSDX[INDEX3(s,2,q,p.row_NN,DIM)]*X_p[INDEX3(k,2,q,p.numEqu,DIM)]);
313 gross 798 }
314 gross 853 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
315     }
316     }
317     } else {
318     for (s=0;s<p.row_NS;s++) {
319     rtmp0=0;
320     rtmp1=0;
321     rtmp2=0;
322     for (q=0;q<p.numQuad;q++) {
323     rtmp0+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_NN,DIM)];
324     rtmp1+=Vol[q]*DSDX[INDEX3(s,1,q,p.row_NN,DIM)];
325     rtmp2+=Vol[q]*DSDX[INDEX3(s,2,q,p.row_NN,DIM)];
326 gross 798 }
327 gross 853 for (k=0;k<p.numEqu;k++) {
328     EM_F[INDEX2(k,s,p.numEqu)]+=rtmp0*X_p[INDEX2(k,0,p.numEqu)]
329     +rtmp1*X_p[INDEX2(k,1,p.numEqu)]
330     +rtmp2*X_p[INDEX2(k,2,p.numEqu)];
331     }
332 gross 798 }
333 gross 853 }
334 gross 798 }
335 gross 853 /**************************************************************/
336     /* process Y: */
337     /**************************************************************/
338 jfenwick 2271 Y_p=getSampleDataRO(Y,e,YBuff);
339 gross 853 if (NULL!=Y_p) {
340     add_EM_F=TRUE;
341     if (extendedY) {
342     for (s=0;s<p.row_NS;s++) {
343     for (k=0;k<p.numEqu;k++) {
344     rtmp=0.;
345     for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*Y_p[INDEX2(k,q,p.numEqu)];
346     EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
347 gross 798 }
348     }
349 gross 853 } else {
350     for (s=0;s<p.row_NS;s++) {
351     rtmp=0;
352     for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)];
353     for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp*Y_p[k];
354 gross 798 }
355 gross 853 }
356     }
357     /***********************************************************************************************/
358     /* add the element matrices onto the matrix and right hand side */
359     /***********************************************************************************************/
360     for (q=0;q<p.row_NN;q++) row_index[q]=p.row_DOF[elements->Nodes[INDEX2(p.row_node[q],e,p.NN)]];
361     if (add_EM_F) Finley_Util_AddScatter(p.row_NN,row_index,p.numEqu,EM_F,F_p, p.row_DOF_UpperBound);
362     if (add_EM_S) Finley_Assemble_addToSystemMatrix(Mat,p.row_NN,row_index,p.numEqu,p.col_NN,row_index,p.numComp,EM_S);
363    
364     } /* end color check */
365     } /* end element loop */
366     } /* end color loop */
367    
368     THREAD_MEMFREE(EM_S);
369     THREAD_MEMFREE(EM_F);
370     THREAD_MEMFREE(row_index);
371 gross 798
372 gross 853 } /* end of pointer check */
373 gross 798 } /* end parallel region */
374 jfenwick 2271 freeSampleBuffer(ABuff);
375     freeSampleBuffer(BBuff);
376     freeSampleBuffer(CBuff);
377     freeSampleBuffer(DBuff);
378     freeSampleBuffer(XBuff);
379     freeSampleBuffer(YBuff);
380 gross 798 }
381     /*
382     * $Log$
383     */

  ViewVC Help
Powered by ViewVC 1.1.26