/[escript]/trunk/dudley/src/Assemble_PDE_Single2_3D.c
ViewVC logotype

Contents of /trunk/dudley/src/Assemble_PDE_Single2_3D.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3223 - (show annotations)
Wed Sep 29 05:02:52 2010 UTC (8 years, 9 months ago) by jfenwick
Original Path: branches/domexper/dudley/src/Assemble_PDE_Single2_3D.c
File MIME type: text/plain
File size: 14048 byte(s)
Remove some - but not all - nasty hacks

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 /* assembles the system of numEq PDEs into the stiffness matrix S right hand side F */
17 /* the shape functions for test and solution must be identical */
18
19 /* -(A_{i,j} u_,j)_i-(B_{i} u)_i+C_{j} u_,j-D u_m and -(X_,i)_i + Y */
20
21 /* in a 3D 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 = 3 x 3 */
27 /* B = 3 */
28 /* C = 3 */
29 /* D = scalar */
30 /* X = 3 */
31 /* Y = scalar */
32
33 /**************************************************************/
34
35 #include "Assemble.h"
36 #include "Util.h"
37 #ifdef _OPENMP
38 #include <omp.h>
39 #endif
40
41 /**************************************************************/
42
43 void Dudley_Assemble_PDE_Single2_3D(Assemble_Parameters p, Dudley_ElementFile * elements,
44 Paso_SystemMatrix * Mat, escriptDataC * F,
45 escriptDataC * A, escriptDataC * B, escriptDataC * C, escriptDataC * D,
46 escriptDataC * X, escriptDataC * Y)
47 {
48
49 #define DIM 3
50 index_t color;
51 dim_t e;
52 __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;
53 double *EM_S, *EM_F, *DSDX;
54 index_t *row_index;
55 register dim_t q, s, r;
56 register double rtmp, rtmp00, rtmp01, rtmp02, rtmp10, rtmp11, rtmp12, rtmp20, rtmp21, rtmp22, rtmp0, rtmp1, rtmp2;
57 bool_t add_EM_F, add_EM_S;
58
59 bool_t extendedA = isExpanded(A);
60 bool_t extendedB = isExpanded(B);
61 bool_t extendedC = isExpanded(C);
62 bool_t extendedD = isExpanded(D);
63 bool_t extendedX = isExpanded(X);
64 bool_t extendedY = isExpanded(Y);
65 double *F_p = (requireWrite(F), getSampleDataRW(F, 0)); /* use comma, to get around the mixed code and declarations thing */
66 const double* S = p.shapeFns;
67 dim_t len_EM_S = p.numShapes * p.numShapes;
68 dim_t len_EM_F = p.numShapes;
69
70 #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,rtmp, rtmp00, rtmp01, rtmp02, rtmp10, rtmp11, rtmp12, rtmp20, rtmp21, rtmp22, rtmp0, rtmp1, rtmp2,add_EM_F, add_EM_S)
71 {
72 EM_S = THREAD_MEMALLOC(len_EM_S, double);
73 EM_F = THREAD_MEMALLOC(len_EM_F, double);
74 row_index = THREAD_MEMALLOC(p.numShapes, index_t);
75
76 if (!Dudley_checkPtr(EM_S) && !Dudley_checkPtr(EM_F) && !Dudley_checkPtr(row_index))
77 {
78
79 for (color = elements->minColor; color <= elements->maxColor; color++)
80 {
81 /* open loop over all elements: */
82 #pragma omp for private(e) schedule(static)
83 for (e = 0; e < elements->numElements; e++)
84 {
85 if (elements->Color[e] == color)
86 {
87
88 A_p = getSampleDataRO(A, e);
89 B_p = getSampleDataRO(B, e);
90 C_p = getSampleDataRO(C, e);
91 D_p = getSampleDataRO(D, e);
92 X_p = getSampleDataRO(X, e);
93 Y_p = getSampleDataRO(Y, e);
94
95 double vol = p.row_jac->absD[e] * p.row_jac->quadweight;
96 DSDX = &(p.row_jac->DSDX[INDEX5(0, 0, 0, 0, e, p.numShapes, DIM, p.numQuad, 1)]);
97 for (q = 0; q < len_EM_S; ++q)
98 EM_S[q] = 0;
99 for (q = 0; q < len_EM_F; ++q)
100 EM_F[q] = 0;
101 add_EM_F = FALSE;
102 add_EM_S = FALSE;
103
104 /**************************************************************/
105 /* process A: */
106 /**************************************************************/
107 if (NULL != A_p)
108 {
109 add_EM_S = TRUE;
110 if (extendedA)
111 {
112 A_q = &(A_p[INDEX4(0, 0, 0, 0, DIM, DIM, p.numQuad)]);
113 for (s = 0; s < p.numShapes; s++)
114 {
115 for (r = 0; r < p.numShapes; r++)
116 {
117 rtmp = 0;
118 for (q = 0; q < p.numQuad; q++)
119 {
120 rtmp +=
121 vol * (DSDX[INDEX3(s, 0, q, p.numShapes, DIM)] *
122 A_q[INDEX3(0, 0, q, DIM, DIM)] *
123 DSDX[INDEX3(r, 0, q, p.numShapes, DIM)] +
124 DSDX[INDEX3(s, 0, q, p.numShapes, DIM)] *
125 A_q[INDEX3(0, 1, q, DIM, DIM)] *
126 DSDX[INDEX3(r, 1, q, p.numShapes, DIM)] +
127 DSDX[INDEX3(s, 0, q, p.numShapes, DIM)] *
128 A_q[INDEX3(0, 2, q, DIM, DIM)] *
129 DSDX[INDEX3(r, 2, q, p.numShapes, DIM)] +
130 DSDX[INDEX3(s, 1, q, p.numShapes, DIM)] *
131 A_q[INDEX3(1, 0, q, DIM, DIM)] *
132 DSDX[INDEX3(r, 0, q, p.numShapes, DIM)] +
133 DSDX[INDEX3(s, 1, q, p.numShapes, DIM)] *
134 A_q[INDEX3(1, 1, q, DIM, DIM)] *
135 DSDX[INDEX3(r, 1, q, p.numShapes, DIM)] +
136 DSDX[INDEX3(s, 1, q, p.numShapes, DIM)] *
137 A_q[INDEX3(1, 2, q, DIM, DIM)] *
138 DSDX[INDEX3(r, 2, q, p.numShapes, DIM)] +
139 DSDX[INDEX3(s, 2, q, p.numShapes, DIM)] *
140 A_q[INDEX3(2, 0, q, DIM, DIM)] *
141 DSDX[INDEX3(r, 0, q, p.numShapes, DIM)] +
142 DSDX[INDEX3(s, 2, q, p.numShapes, DIM)] *
143 A_q[INDEX3(2, 1, q, DIM, DIM)] *
144 DSDX[INDEX3(r, 1, q, p.numShapes, DIM)] +
145 DSDX[INDEX3(s, 2, q, p.numShapes, DIM)] *
146 A_q[INDEX3(2, 2, q, DIM, DIM)] *
147 DSDX[INDEX3(r, 2, q, p.numShapes, DIM)]);
148 }
149 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] += rtmp;
150 }
151 }
152 } else
153 {
154 for (s = 0; s < p.numShapes; s++)
155 {
156 for (r = 0; r < p.numShapes; r++)
157 {
158 rtmp00 = 0;
159 rtmp01 = 0;
160 rtmp02 = 0;
161 rtmp10 = 0;
162 rtmp11 = 0;
163 rtmp12 = 0;
164 rtmp20 = 0;
165 rtmp21 = 0;
166 rtmp22 = 0;
167 for (q = 0; q < p.numQuad; q++)
168 {
169
170 rtmp0 = vol * DSDX[INDEX3(s, 0, q, p.numShapes, DIM)];
171 rtmp00 += rtmp0 * DSDX[INDEX3(r, 0, q, p.numShapes, DIM)];
172 rtmp01 += rtmp0 * DSDX[INDEX3(r, 1, q, p.numShapes, DIM)];
173 rtmp02 += rtmp0 * DSDX[INDEX3(r, 2, q, p.numShapes, DIM)];
174
175 rtmp1 = vol * DSDX[INDEX3(s, 1, q, p.numShapes, DIM)];
176 rtmp10 += rtmp1 * DSDX[INDEX3(r, 0, q, p.numShapes, DIM)];
177 rtmp11 += rtmp1 * DSDX[INDEX3(r, 1, q, p.numShapes, DIM)];
178 rtmp12 += rtmp1 * DSDX[INDEX3(r, 2, q, p.numShapes, DIM)];
179
180 rtmp2 = vol * DSDX[INDEX3(s, 2, q, p.numShapes, DIM)];
181 rtmp20 += rtmp2 * DSDX[INDEX3(r, 0, q, p.numShapes, DIM)];
182 rtmp21 += rtmp2 * DSDX[INDEX3(r, 1, q, p.numShapes, DIM)];
183 rtmp22 += rtmp2 * DSDX[INDEX3(r, 2, q, p.numShapes, DIM)];
184 }
185 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] +=
186 rtmp00 * A_p[INDEX2(0, 0, DIM)] + rtmp01 * A_p[INDEX2(0, 1, DIM)] +
187 rtmp02 * A_p[INDEX2(0, 2, DIM)] + rtmp10 * A_p[INDEX2(1, 0, DIM)] +
188 rtmp11 * A_p[INDEX2(1, 1, DIM)] + rtmp12 * A_p[INDEX2(1, 2, DIM)] +
189 rtmp20 * A_p[INDEX2(2, 0, DIM)] + rtmp21 * A_p[INDEX2(2, 1, DIM)] +
190 rtmp22 * A_p[INDEX2(2, 2, DIM)];
191 }
192 }
193 }
194 }
195 /**************************************************************/
196 /* process B: */
197 /**************************************************************/
198 if (NULL != B_p)
199 {
200 add_EM_S = TRUE;
201 if (extendedB)
202 {
203 B_q = &(B_p[INDEX3(0, 0, 0, DIM, p.numQuad)]);
204 for (s = 0; s < p.numShapes; s++)
205 {
206 for (r = 0; r < p.numShapes; r++)
207 {
208 rtmp = 0;
209 for (q = 0; q < p.numQuad; q++)
210 {
211 rtmp += vol * S[INDEX2(r, q, p.numShapes)] *
212 (DSDX[INDEX3(s, 0, q, p.numShapes, DIM)] *
213 B_q[INDEX2(0, q, DIM)] +
214 DSDX[INDEX3(s, 1, q, p.numShapes, DIM)] *
215 B_q[INDEX2(1, q, DIM)] +
216 DSDX[INDEX3(s, 2, q, p.numShapes, DIM)] *
217 B_q[INDEX2(2, q, DIM)]);
218 }
219 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] += rtmp;
220 }
221 }
222 } else
223 {
224 for (s = 0; s < p.numShapes; s++)
225 {
226 for (r = 0; r < p.numShapes; r++)
227 {
228 rtmp0 = 0;
229 rtmp1 = 0;
230 rtmp2 = 0;
231 for (q = 0; q < p.numQuad; q++)
232 {
233 rtmp = vol * S[INDEX2(r, q, p.numShapes)];
234 rtmp0 += rtmp * DSDX[INDEX3(s, 0, q, p.numShapes, DIM)];
235 rtmp1 += rtmp * DSDX[INDEX3(s, 1, q, p.numShapes, DIM)];
236 rtmp2 += rtmp * DSDX[INDEX3(s, 2, q, p.numShapes, DIM)];
237 }
238 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] +=
239 rtmp0 * B_p[0] + rtmp1 * B_p[1] + rtmp2 * B_p[2];
240 }
241 }
242 }
243 }
244 /**************************************************************/
245 /* process C: */
246 /**************************************************************/
247 if (NULL != C_p)
248 {
249 add_EM_S = TRUE;
250 if (extendedC)
251 {
252 C_q = &(C_p[INDEX3(0, 0, 0, DIM, p.numQuad)]);
253 for (s = 0; s < p.numShapes; s++)
254 {
255 for (r = 0; r < p.numShapes; r++)
256 {
257 rtmp = 0;
258 for (q = 0; q < p.numQuad; q++)
259 {
260 rtmp += vol * S[INDEX2(s, q, p.numShapes)] *
261 (C_q[INDEX2(0, q, DIM)] *
262 DSDX[INDEX3(r, 0, q, p.numShapes, DIM)] +
263 C_q[INDEX2(1, q, DIM)] *
264 DSDX[INDEX3(r, 1, q, p.numShapes, DIM)] +
265 C_q[INDEX2(2, q, DIM)] *
266 DSDX[INDEX3(r, 2, q, p.numShapes, DIM)]);
267 }
268 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] += rtmp;
269 }
270 }
271 } else
272 {
273 for (s = 0; s < p.numShapes; s++)
274 {
275 for (r = 0; r < p.numShapes; r++)
276 {
277 rtmp0 = 0;
278 rtmp1 = 0;
279 rtmp2 = 0;
280 for (q = 0; q < p.numQuad; q++)
281 {
282 rtmp = vol * S[INDEX2(s, q, p.numShapes)];
283 rtmp0 += rtmp * DSDX[INDEX3(r, 0, q, p.numShapes, DIM)];
284 rtmp1 += rtmp * DSDX[INDEX3(r, 1, q, p.numShapes, DIM)];
285 rtmp2 += rtmp * DSDX[INDEX3(r, 2, q, p.numShapes, DIM)];
286 }
287 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] +=
288 rtmp0 * C_p[0] + rtmp1 * C_p[1] + rtmp2 * C_p[2];
289 }
290 }
291 }
292 }
293 /************************************************************* */
294 /* process D */
295 /**************************************************************/
296 if (NULL != D_p)
297 {
298 add_EM_S = TRUE;
299 if (extendedD)
300 {
301 D_q = &(D_p[INDEX2(0, 0, p.numQuad)]);
302 for (s = 0; s < p.numShapes; s++)
303 {
304 for (r = 0; r < p.numShapes; r++)
305 {
306 rtmp = 0;
307 for (q = 0; q < p.numQuad; q++)
308 rtmp +=
309 vol * S[INDEX2(s, q, p.numShapes)] * D_q[q] *
310 S[INDEX2(r, q, p.numShapes)];
311 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] += rtmp;
312 }
313 }
314 } else
315 {
316 for (s = 0; s < p.numShapes; s++)
317 {
318 for (r = 0; r < p.numShapes; r++)
319 {
320 rtmp = 0;
321 for (q = 0; q < p.numQuad; q++)
322 rtmp +=
323 vol * S[INDEX2(s, q, p.numShapes)] *
324 S[INDEX2(r, q, p.numShapes)];
325 EM_S[INDEX4(0, 0, s, r, p.numEqu, p.numComp, p.numShapes)] +=
326 rtmp * D_p[0];
327 }
328 }
329 }
330 }
331 /**************************************************************/
332 /* process X: */
333 /**************************************************************/
334 if (NULL != X_p)
335 {
336 add_EM_F = TRUE;
337 if (extendedX)
338 {
339 X_q = &(X_p[INDEX3(0, 0, 0, DIM, p.numQuad)]);
340 for (s = 0; s < p.numShapes; s++)
341 {
342 rtmp = 0;
343 for (q = 0; q < p.numQuad; q++)
344 {
345 rtmp +=
346 vol * (DSDX[INDEX3(s, 0, q, p.numShapes, DIM)] *
347 X_q[INDEX2(0, q, DIM)] +
348 DSDX[INDEX3(s, 1, q, p.numShapes, DIM)] *
349 X_q[INDEX2(1, q, DIM)] +
350 DSDX[INDEX3(s, 2, q, p.numShapes, DIM)] *
351 X_q[INDEX2(2, q, DIM)]);
352 }
353 EM_F[INDEX2(0, s, p.numEqu)] += rtmp;
354 }
355 } else
356 {
357 for (s = 0; s < p.numShapes; s++)
358 {
359 rtmp0 = 0;
360 rtmp1 = 0;
361 rtmp2 = 0;
362 for (q = 0; q < p.numQuad; q++)
363 {
364 rtmp0 += vol * DSDX[INDEX3(s, 0, q, p.numShapes, DIM)];
365 rtmp1 += vol * DSDX[INDEX3(s, 1, q, p.numShapes, DIM)];
366 rtmp2 += vol * DSDX[INDEX3(s, 2, q, p.numShapes, DIM)];
367 }
368 EM_F[INDEX2(0, s, p.numEqu)] += rtmp0 * X_p[0] + rtmp1 * X_p[1] + rtmp2 * X_p[2];
369 }
370 }
371 }
372 /**************************************************************/
373 /* process Y: */
374 /**************************************************************/
375 if (NULL != Y_p)
376 {
377 add_EM_F = TRUE;
378 if (extendedY)
379 {
380 Y_q = &(Y_p[INDEX2(0, 0, p.numQuad)]);
381 for (s = 0; s < p.numShapes; s++)
382 {
383 rtmp = 0;
384 for (q = 0; q < p.numQuad; q++)
385 rtmp += vol * S[INDEX2(s, q, p.numShapes)] * Y_q[q];
386 EM_F[INDEX2(0, s, p.numEqu)] += rtmp;
387 }
388 } else
389 {
390 for (s = 0; s < p.numShapes; s++)
391 {
392 rtmp = 0;
393 for (q = 0; q < p.numQuad; q++)
394 rtmp += vol * S[INDEX2(s, q, p.numShapes)];
395 EM_F[INDEX2(0, s, p.numEqu)] += rtmp * Y_p[0];
396 }
397 }
398 }
399 /***********************************************************************************************/
400 /* add the element matrices onto the matrix and right hand side */
401 /***********************************************************************************************/
402
403 for (q = 0; q < p.numShapes; q++)
404 row_index[q] = p.row_DOF[elements->Nodes[INDEX2(q, e, p.NN)]];
405
406 if (add_EM_F)
407 Dudley_Util_AddScatter(p.numShapes, row_index, p.numEqu, EM_F, F_p,
408 p.row_DOF_UpperBound);
409 if (add_EM_S)
410 Dudley_Assemble_addToSystemMatrix(Mat, p.numShapes, row_index, p.numEqu,
411 p.numShapes, row_index, p.numComp, EM_S);
412
413 } /* end color check */
414 } /* end element loop */
415 } /* end color loop */
416
417 THREAD_MEMFREE(EM_S);
418 THREAD_MEMFREE(EM_F);
419 THREAD_MEMFREE(row_index);
420
421 } /* end of pointer check */
422 } /* end parallel region */
423 }

  ViewVC Help
Powered by ViewVC 1.1.26