ZeroOrderAssembler.cc 12 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
#include <vector>
#include "Assembler.h"
#include "ZeroOrderAssembler.h"
#include "Operator.h"
#include "QPsiPhi.h"
#include "FiniteElemSpace.h"
#include "Quadrature.h"
#include "DOFVector.h"
#include "ElementMatrix.h"
#include "OpenMP.h"

namespace AMDiS {

  std::vector<SubAssembler*> ZeroOrderAssembler::optimizedSubAssemblers;
  std::vector<SubAssembler*> ZeroOrderAssembler::standardSubAssemblers;

  ZeroOrderAssembler::ZeroOrderAssembler(Operator *op,
					 Assembler *assembler,
					 Quadrature *quad,
					 bool optimized)
    : SubAssembler(op, assembler, quad, 0, optimized)
  {}

  ZeroOrderAssembler* ZeroOrderAssembler::getSubAssembler(Operator* op,
							  Assembler *assembler,
							  Quadrature *quad,
							  bool optimized)
  {
    // check if a assembler is needed at all
    if (op->zeroOrder.size() == 0) {
      return NULL;
    }

    ZeroOrderAssembler *newAssembler;

    std::vector<SubAssembler*> *subAssemblers =
37
      optimized ? &optimizedSubAssemblers : &standardSubAssemblers;
38
39

    int myRank = omp_get_thread_num();
40
    std::vector<OperatorTerm*> opTerms = op->zeroOrder[myRank];
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149

    sort(opTerms.begin(), opTerms.end());

    // check if a new assembler is needed
    if (quad) {
      for (int i = 0; i < static_cast<int>( subAssemblers->size()); i++) {
	std::vector<OperatorTerm*> assTerms = *((*subAssemblers)[i]->getTerms());

	sort(assTerms.begin(), assTerms.end());

	if ((opTerms == assTerms) && 
	    ((*subAssemblers)[i]->getQuadrature() == quad)) {
	
	  return dynamic_cast<ZeroOrderAssembler*>((*subAssemblers)[i]);
	}
      }
    }
 
    // check if all terms are pw_const
    bool pwConst = true;
    for (int i = 0; i < static_cast<int>( op->zeroOrder[myRank].size()); i++) {
      if (!op->zeroOrder[myRank][i]->isPWConst()) {
	pwConst = false;
	break;
      }
    }  

    // create new assembler
    if (!optimized) {
      newAssembler = NEW StandardZOA(op, assembler, quad);
    } else {
      if (pwConst) {
	newAssembler = NEW PrecalcZOA(op, assembler, quad);
      } else {
	newAssembler = NEW FastQuadZOA(op, assembler, quad);
      }
    }

    subAssemblers->push_back(newAssembler);
    return newAssembler;
  }

  StandardZOA::StandardZOA(Operator *op, Assembler *assembler, Quadrature *quad)
    : ZeroOrderAssembler(op, assembler, quad, false)
  {
  }

  void StandardZOA::calculateElementMatrix(const ElInfo *elInfo, ElementMatrix *mat)
  {
    const BasisFunction *psi = owner->getRowFESpace()->getBasisFcts();
    const BasisFunction *phi = owner->getColFESpace()->getBasisFcts();

    double *phival = GET_MEMORY(double, nCol);
    int nPoints = quadrature->getNumPoints();
    double *c = GET_MEMORY(double, nPoints);

    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] = 0.0;
    }

    int myRank = omp_get_thread_num();
    std::vector<OperatorTerm*>::iterator termIt;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>((*termIt)))->getC(elInfo, nPoints, c);
    }
      
    if (symmetric) {
      for (int iq = 0; iq < nPoints; iq++) {
	c[iq] *= elInfo->getDet();

	// calculate phi at QPs only once!
	for (int i = 0; i < nCol; i++) {
	  phival[i] = (*(phi->getPhi(i)))(quadrature->getLambda(iq));
	}

	for (int i = 0; i < nRow; i++) {
	  double psival = (*(psi->getPhi(i)))(quadrature->getLambda(iq));
	  (*mat)[i][i] += quadrature->getWeight(iq) * c[iq] * psival * phival[i];
	  for (int j = i + 1; j < nCol; j++) {
	    double val = quadrature->getWeight(iq) * c[iq] * psival * phival[j];
	    (*mat)[i][j] += val;
	    (*mat)[j][i] += val;
	  }
	}
      }
    } else {      //  non symmetric assembling 
      for (int iq = 0; iq < nPoints; iq++) {
	c[iq] *= elInfo->getDet();

	// calculate phi at QPs only once!
	for (int i = 0; i < nCol; i++) {
	  phival[i] = (*(phi->getPhi(i)))(quadrature->getLambda(iq));
	}

	for (int i = 0; i < nRow; i++) {
	  double psival = (*(psi->getPhi(i)))(quadrature->getLambda(iq));
	  for (int j = 0; j < nCol; j++) {
	    (*mat)[i][j] += quadrature->getWeight(iq) * c[iq] * psival * phival[j];
	  }
	}
      }
    }

    FREE_MEMORY(phival, double, nCol);
    FREE_MEMORY(c, double, nPoints);
  }

  void StandardZOA::calculateElementMatrix(const ElInfo *rowElInfo,
					   const ElInfo *colElInfo,
150
151
					   const ElInfo *smallElInfo,
					   const ElInfo *largeElInfo,
152
153
					   ElementMatrix *mat) 
  {
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
    FUNCNAME("StandardZOA::calculateElementMatrix()");

    TEST_EXIT((nRow <= 3) && (nCol <= 3))("not yet!\n");

    const BasisFunction *psi = owner->getRowFESpace()->getBasisFcts();
    const BasisFunction *phi = owner->getColFESpace()->getBasisFcts();
    DimMat<double> *m = smallElInfo->getSubElemCoordsMat();
    //    m->print();
    //    WAIT_REALLY;

    // The basis function on the larger element is defined as a linear
    // combination of basis functions of the smaller element. At the moment,
    // this is supported only for linear lagrange basis functions.
    //
    // If b0 and b1 are the basis functions of the larger element restricted to 
    // the small element, and s0 and s1 are the two basis function of the small 
    // element, b0 and b1 are expressed as follows:
    //
    // b0 = c00 * s0 + c01 * s1;
    // b1 = c10 * s0 + c11 * s1;
    //
    // The constants are defined by the subElement Matrix.

177
178
179
180

    int nPoints = quadrature->getNumPoints();
    double *c = GET_MEMORY(double, nPoints);

181
182
183
184
    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] = 0.0;
    }

185
186
187
188
189
    int myRank = omp_get_thread_num();
    std::vector<OperatorTerm*>::iterator termIt;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>((*termIt)))->getC(rowElInfo, nPoints, c);
    }
190
    
191
    for (int iq = 0; iq < nPoints; iq++) {
192
      c[iq] *= smallElInfo->getDet();
193

194
195
196
      for (int i = 0; i < nCol; i++) { 
	for (int j = 0; j < nRow; j++) { 
	  double val = quadrature->getWeight(iq) * c[iq] * (*(phi->getPhi(i)))(quadrature->getLambda(iq));
197

198
199
200
	  double tmpval = 0.0;
	  for (int k = 0; k < nCol; k++) {
	    tmpval += (*m)[j][k] * (*(phi->getPhi(k)))(quadrature->getLambda(iq));
201
	  }
202
203
204
	  val *= tmpval;

	  (*mat)[j][i] += val;
205
206
	}
      }
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
    }
  }

  void StandardZOA::calculateElementVector(const ElInfo *elInfo, ElementVector *vec)
  {
    int nPoints = quadrature->getNumPoints();

    double *c = GET_MEMORY(double, nPoints);

    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] = 0.0;
    }

    int myRank = omp_get_thread_num();
    std::vector<OperatorTerm*>::iterator termIt;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>((*termIt)))->getC(elInfo, nPoints, c);
    }

    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] *= elInfo->getDet();

      for (int i = 0; i < nRow; i++) {
	double psi = (*(owner->getRowFESpace()->getBasisFcts()->getPhi(i)))
	  (quadrature->getLambda(iq));
	(*vec)[i] += quadrature->getWeight(iq) * c[iq] * psi;
      }
    }
    
    FREE_MEMORY(c, double, nPoints);
  }

  FastQuadZOA::FastQuadZOA(Operator *op, Assembler *assembler, Quadrature *quad)
    : ZeroOrderAssembler(op, assembler, quad, true)
  {
242
    cPtrs.resize(omp_get_overall_max_threads());
243
244
245
246
  }

  FastQuadZOA::~FastQuadZOA()
  {
247
    for (int i = 0; i < omp_get_overall_max_threads(); i++) {
248
249
250
251
252
253
254
255
256
257
      FREE_MEMORY(cPtrs[i], double, quadrature->getNumPoints());
    }
  }

  void FastQuadZOA::calculateElementMatrix(const ElInfo *elInfo, ElementMatrix *mat)
  {
    int nPoints = quadrature->getNumPoints();
    int myRank = omp_get_thread_num();

    if (firstCall) {
258
259
260
261
262
263
264
265
266
267
268
#ifdef _OPENMP
#pragma omp critical
#endif 
      {
	cPtrs[myRank] = GET_MEMORY(double, nPoints);
	const BasisFunction *basFcts = owner->getRowFESpace()->getBasisFcts();
	psiFast = updateFastQuadrature(psiFast, basFcts, INIT_PHI);
	basFcts = owner->getColFESpace()->getBasisFcts();
	phiFast = updateFastQuadrature(phiFast, basFcts, INIT_PHI);
	firstCall = false;
      }
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
    }

    double *c = cPtrs[myRank];
    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] = 0.0;
    }

    std::vector<OperatorTerm*>::iterator termIt;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>((*termIt)))->getC(elInfo, nPoints, c);
    }

    if (symmetric) {
      for (int iq = 0; iq < nPoints; iq++) {
	c[iq] *= elInfo->getDet();

	const double *psi = psiFast->getPhi(iq);
	const double *phi = phiFast->getPhi(iq);
	for (int i = 0; i < nRow; i++) {
	  (*mat)[i][i] += quadrature->getWeight(iq) * c[iq] * psi[i] * phi[i];
	  for (int j = i + 1; j < nCol; j++) {
	    double val = quadrature->getWeight(iq) * c[iq] * psi[i] * phi[j];
	    (*mat)[i][j] += val;
	    (*mat)[j][i] += val;
	  }
	}
      }
    } else {      /*  non symmetric assembling   */
      for (int iq = 0; iq < nPoints; iq++) {
	c[iq] *= elInfo->getDet();

	const double *psi = psiFast->getPhi(iq);
	const double *phi = phiFast->getPhi(iq);
	for (int i = 0; i < nRow; i++) {
	  for (int j = 0; j < nCol; j++) {
	    (*mat)[i][j] += quadrature->getWeight(iq) * c[iq] * psi[i] * phi[j];
	  }
	}
      }
    }
  }

  void FastQuadZOA::calculateElementVector(const ElInfo *elInfo, ElementVector *vec)
  {
    if (firstCall) {
314
315
316
317
318
319
320
321
322
323
#ifdef _OPENMP
#pragma omp critical
#endif 
      {
	const BasisFunction *basFcts = owner->getRowFESpace()->getBasisFcts();
	psiFast = updateFastQuadrature(psiFast, basFcts, INIT_PHI);
	basFcts = owner->getColFESpace()->getBasisFcts();
	phiFast = updateFastQuadrature(phiFast, basFcts, INIT_PHI);
	firstCall = false;
      }
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
    }

    int nPoints = quadrature->getNumPoints();
    double *c = GET_MEMORY(double, nPoints);

    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] = 0.0;
    }

    int myRank = omp_get_thread_num();
    std::vector<OperatorTerm*>::iterator termIt;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>((*termIt)))->getC(elInfo, nPoints, c);
    }

    for (int iq = 0; iq < nPoints; iq++) {
      c[iq] *= elInfo->getDet();

      const double *psi = psiFast->getPhi(iq);
      for (int i = 0; i < nRow; i++) {
	(*vec)[i] += quadrature->getWeight(iq) * c[iq] * psi[i];
      }
    }
    FREE_MEMORY(c, double, nPoints);
  }

  PrecalcZOA::PrecalcZOA(Operator *op, Assembler *assembler, Quadrature *quad) 
    : ZeroOrderAssembler(op, assembler, quad, true)
  {
  }

  void PrecalcZOA::calculateElementMatrix(const ElInfo *elInfo, ElementMatrix *mat)
  {
    if (firstCall) {
358
359
360
361
362
363
364
365
366
367
368
#ifdef _OPENMP
#pragma omp critical
#endif 
      {
	q00 = Q00PsiPhi::provideQ00PsiPhi(owner->getRowFESpace()->getBasisFcts(), 
					  owner->getColFESpace()->getBasisFcts(), 
					  quadrature);
	q0 = Q0Psi::provideQ0Psi(owner->getRowFESpace()->getBasisFcts(),
				 quadrature);
	firstCall = false;
      }
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
    }

    double c = 0.0;
    int myRank = omp_get_thread_num();
    int size = static_cast<int>(terms[myRank].size());

    for (int i = 0; i < size; i++) {
      (static_cast<ZeroOrderTerm*>((terms[myRank][i])))->getC(elInfo, 1, &c);
    }

    c *= elInfo->getDet();

    if (symmetric) {
      for (int i = 0; i < nRow; i++) {
	(*mat)[i][i] += c * q00->getValue(i,i);
	for (int j = i + 1; j < nCol; j++) {
	  double val = c * q00->getValue(i, j);
	  (*mat)[i][j] += val;
	  (*mat)[j][i] += val;
	}
      }
    } else {
      for (int i = 0; i < nRow; i++)
	for (int j = 0; j < nCol; j++)
	  (*mat)[i][j] += c * q00->getValue(i, j);
    }
  }

  void PrecalcZOA::calculateElementVector(const ElInfo *elInfo, ElementVector *vec)
  {
    if (firstCall) {
400
401
402
403
404
405
406
407
408
409
410
#ifdef _OPENMP
#pragma omp critical
#endif 
      {
	q00 = Q00PsiPhi::provideQ00PsiPhi(owner->getRowFESpace()->getBasisFcts(), 
					  owner->getColFESpace()->getBasisFcts(), 
					  quadrature);
	q0 = Q0Psi::provideQ0Psi(owner->getRowFESpace()->getBasisFcts(),
				 quadrature);	
	firstCall = false;
      }
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
    }

    std::vector<OperatorTerm*>::iterator termIt;

    int myRank = omp_get_thread_num();
    double c = 0.0;
    for (termIt = terms[myRank].begin(); termIt != terms[myRank].end(); ++termIt) {
      (static_cast<ZeroOrderTerm*>( *termIt))->getC(elInfo, 1, &c);
    }

    c *= elInfo->getDet();

    for (int i = 0; i < nRow; i++)
      (*vec)[i] += c * q0->getValue(i);
  }

}