/[escript]/trunk/finley/src/Assemble_PDE_System2_1D.c
ViewVC logotype

Contents of /trunk/finley/src/Assemble_PDE_System2_1D.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 2770 - (show annotations)
Wed Nov 25 01:24:51 2009 UTC (10 years ago) by jfenwick
File MIME type: text/plain
File size: 16152 byte(s)
Removed buffer implementation of Lazy.
Removed supporting Alloc/Free Sample buffer calls.
1
2 /*******************************************************
3 *
4 * Copyright (c) 2003-2009 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 1D 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 1 x p.numComp x 1 */
29 /* B = 1 x numEqu x p.numComp */
30 /* C = p.numEqu x 1 x p.numComp */
31 /* D = p.numEqu x p.numComp */
32 /* X = p.numEqu x 1 */
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_1D(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 1
53 index_t color;
54 dim_t e, isub;
55 __const double *A_p, *B_p, *C_p, *D_p, *X_p, *Y_p, *A_q, *B_q, *C_q, *D_q, *X_q, *Y_q;
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;
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->BasisFunctions->S;
70 dim_t len_EM_S=p.row_numShapesTotal*p.col_numShapesTotal*p.numEqu*p.numComp;
71 dim_t len_EM_F=p.row_numShapesTotal*p.numEqu;
72
73 #pragma omp parallel private(color, EM_S, EM_F, Vol, DSDX, A_p, B_p, C_p, D_p, X_p, Y_p, A_q, B_q, C_q, D_q, X_q, Y_q,row_index, q, s,r,k,m,rtmp,add_EM_F, add_EM_S, isub)
74 {
75 EM_S=THREAD_MEMALLOC(len_EM_S,double);
76 EM_F=THREAD_MEMALLOC(len_EM_F,double);
77 row_index=THREAD_MEMALLOC(p.row_numShapesTotal,index_t);
78
79 if (!Finley_checkPtr(EM_S) && !Finley_checkPtr(EM_F) && !Finley_checkPtr(row_index) ) {
80
81 for (color=elements->minColor;color<=elements->maxColor;color++) {
82 /* open loop over all elements: */
83 #pragma omp for private(e) schedule(static)
84 for(e=0;e<elements->numElements;e++){
85 if (elements->Color[e]==color) {
86
87 A_p=getSampleDataRO(A,e);
88 B_p=getSampleDataRO(B,e);
89 C_p=getSampleDataRO(C,e);
90 D_p=getSampleDataRO(D,e);
91 X_p=getSampleDataRO(X,e);
92 Y_p=getSampleDataRO(Y,e);
93
94 for (isub=0; isub<p.numSub; isub++) {
95
96 Vol=&(p.row_jac->volume[INDEX3(0,isub,e, p.numQuadSub,p.numSub)]);
97 DSDX=&(p.row_jac->DSDX[INDEX5(0,0,0,isub,e, p.row_numShapesTotal,DIM,p.numQuadSub,p.numSub)]);
98 for (q=0;q<len_EM_S;++q) EM_S[q]=0;
99 for (q=0;q<len_EM_F;++q) EM_F[q]=0;
100 add_EM_F=FALSE;
101 add_EM_S=FALSE;
102
103 /**************************************************************/
104 /* process A: */
105 /**************************************************************/
106
107 if (NULL!=A_p) {
108 add_EM_S=TRUE;
109 if (extendedA) {
110 A_q=&(A_p[INDEX6(0,0,0,0,0,isub, p.numEqu,DIM,p.numComp,DIM,p.numQuadSub)]);
111 for (s=0;s<p.row_numShapes;s++) {
112 for (r=0;r<p.col_numShapes;r++) {
113 for (k=0;k<p.numEqu;k++) {
114 for (m=0;m<p.numComp;m++) {
115 rtmp=0.;
116 for (q=0;q<p.numQuadSub;q++) {
117 rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)]*
118 A_q[INDEX5(k,0,m,0,q,p.numEqu,DIM,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_numShapesTotal,DIM)];
119 }
120 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp;
121 }
122 }
123 }
124 }
125 } else {
126 for (s=0;s<p.row_numShapes;s++) {
127 for (r=0;r<p.col_numShapes;r++) {
128 rtmp=0;
129 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)]*DSDX[INDEX3(r,0,q,p.row_numShapesTotal,DIM)];
130 for (k=0;k<p.numEqu;k++) {
131 for (m=0;m<p.numComp;m++) {
132 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp*A_p[INDEX4(k,0,m,0,p.numEqu,DIM,p.numComp)];
133 }
134 }
135 }
136 }
137 }
138 }
139 /**************************************************************/
140 /* process B: */
141 /**************************************************************/
142
143 if (NULL!=B_p) {
144 add_EM_S=TRUE;
145 if (extendedB) {
146 B_q=&(B_p[INDEX5(0,0,0,0,isub, p.numEqu,DIM,p.numComp,p.numQuadSub)]);
147 for (s=0;s<p.row_numShapes;s++) {
148 for (r=0;r<p.col_numShapes;r++) {
149 for (k=0;k<p.numEqu;k++) {
150 for (m=0;m<p.numComp;m++) {
151 rtmp=0.;
152 for (q=0;q<p.numQuadSub;q++) {
153 rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)]*
154 B_q[INDEX4(k,0,m,q,p.numEqu,DIM,p.numComp)]*S[INDEX2(r,q,p.row_numShapes)];
155 }
156 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp;
157 }
158 }
159 }
160 }
161 } else {
162 for (s=0;s<p.row_numShapes;s++) {
163 for (r=0;r<p.col_numShapes;r++) {
164 rtmp=0;
165 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)]*S[INDEX2(r,q,p.row_numShapes)];
166 for (k=0;k<p.numEqu;k++) {
167 for (m=0;m<p.numComp;m++) {
168 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp*B_p[INDEX3(k,0,m,p.numEqu,DIM)];
169 }
170 }
171 }
172 }
173 }
174 }
175 /**************************************************************/
176 /* process C: */
177 /**************************************************************/
178
179 if (NULL!=C_p) {
180 add_EM_S=TRUE;
181 if (extendedC) {
182 C_q=&(C_p[INDEX5(0,0,0,0,isub, p.numEqu,p.numComp,DIM, p.numQuadSub)]);
183 for (s=0;s<p.row_numShapes;s++) {
184 for (r=0;r<p.col_numShapes;r++) {
185 for (k=0;k<p.numEqu;k++) {
186 for (m=0;m<p.numComp;m++) {
187 rtmp=0;
188 for (q=0;q<p.numQuadSub;q++) {
189 rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)]*C_q[INDEX4(k,m,0,q,p.numEqu,p.numComp,DIM)]*DSDX[INDEX3(r,0,q,p.row_numShapesTotal,DIM)];
190 }
191 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp;
192 }
193 }
194 }
195 }
196 } else {
197 for (s=0;s<p.row_numShapes;s++) {
198 for (r=0;r<p.col_numShapes;r++) {
199 rtmp=0;
200 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)]*DSDX[INDEX3(r,0,q,p.row_numShapesTotal,DIM)];
201 for (k=0;k<p.numEqu;k++) {
202 for (m=0;m<p.numComp;m++) {
203 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp*C_p[INDEX3(k,m,0,p.numEqu,p.numComp)];
204 }
205 }
206 }
207 }
208 }
209 }
210 /************************************************************* */
211 /* process D */
212 /**************************************************************/
213
214 if (NULL!=D_p) {
215 add_EM_S=TRUE;
216 if (extendedD) {
217 D_q=&(D_p[INDEX4(0,0,0,isub, p.numEqu,p.numComp,p.numQuadSub)]);
218 for (s=0;s<p.row_numShapes;s++) {
219 for (r=0;r<p.col_numShapes;r++) {
220 for (k=0;k<p.numEqu;k++) {
221 for (m=0;m<p.numComp;m++) {
222 rtmp=0;
223 for (q=0;q<p.numQuadSub;q++) {
224 rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)]*D_q[INDEX3(k,m,q,p.numEqu,p.numComp)]*S[INDEX2(r,q,p.row_numShapes)];
225 }
226 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp;
227
228 }
229 }
230 }
231 }
232 } else {
233 for (s=0;s<p.row_numShapes;s++) {
234 for (r=0;r<p.col_numShapes;r++) {
235 rtmp=0;
236 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)]*S[INDEX2(r,q,p.row_numShapes)];
237 for (k=0;k<p.numEqu;k++) {
238 for (m=0;m<p.numComp;m++) {
239 EM_S[INDEX4(k,m,s,r,p.numEqu,p.numComp,p.row_numShapesTotal)]+=rtmp*D_p[INDEX2(k,m,p.numEqu)];
240 }
241 }
242 }
243 }
244 }
245 }
246 /**************************************************************/
247 /* process X: */
248 /**************************************************************/
249
250 if (NULL!=X_p) {
251 add_EM_F=TRUE;
252 if (extendedX) {
253 X_q=&(X_p[INDEX4(0,0,0,isub, p.numEqu,DIM,p.numQuadSub)]);
254 for (s=0;s<p.row_numShapes;s++) {
255 for (k=0;k<p.numEqu;k++) {
256 rtmp=0;
257 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)]*X_q[INDEX3(k,0,q,p.numEqu,DIM)];
258 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
259 }
260 }
261 } else {
262 for (s=0;s<p.row_numShapes;s++) {
263 rtmp=0;
264 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*DSDX[INDEX3(s,0,q,p.row_numShapesTotal,DIM)];
265 for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp*X_p[INDEX2(k,0,p.numEqu)];
266 }
267 }
268 }
269 /**************************************************************/
270 /* process Y: */
271 /**************************************************************/
272
273 if (NULL!=Y_p) {
274 add_EM_F=TRUE;
275 if (extendedY) {
276 Y_q=&(Y_p[INDEX3(0,0,isub, p.numEqu,p.numQuadSub)]);
277 for (s=0;s<p.row_numShapes;s++) {
278 for (k=0;k<p.numEqu;k++) {
279 rtmp=0;
280 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)]*Y_q[INDEX2(k,q,p.numEqu)];
281 EM_F[INDEX2(k,s,p.numEqu)]+=rtmp;
282 }
283 }
284 } else {
285 for (s=0;s<p.row_numShapes;s++) {
286 rtmp=0;
287 for (q=0;q<p.numQuadSub;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_numShapes)];
288 for (k=0;k<p.numEqu;k++) EM_F[INDEX2(k,s,p.numEqu)]+=rtmp*Y_p[k];
289 }
290 }
291 }
292 /***********************************************************************************************/
293 /* add the element matrices onto the matrix and right hand side */
294 /***********************************************************************************************/
295
296 for (q=0;q<p.row_numShapesTotal;q++) row_index[q]=p.row_DOF[elements->Nodes[INDEX2(p.row_node[INDEX2(q,isub,p.row_numShapesTotal)],e,p.NN)]];
297
298 if (add_EM_F) Finley_Util_AddScatter(p.row_numShapesTotal,row_index,p.numEqu,EM_F,F_p, p.row_DOF_UpperBound);
299 if (add_EM_S) Finley_Assemble_addToSystemMatrix(Mat,p.row_numShapesTotal,row_index,p.numEqu,p.col_numShapesTotal,row_index,p.numComp,EM_S);
300
301 } /* end of isub */
302 } /* end color check */
303 } /* end element loop */
304 } /* end color loop */
305
306 THREAD_MEMFREE(EM_S);
307 THREAD_MEMFREE(EM_F);
308 THREAD_MEMFREE(row_index);
309
310 } /* end of pointer check */
311 } /* end parallel region */
312 }
313 /*
314 * $Log$
315 */

  ViewVC Help
Powered by ViewVC 1.1.26