/[escript]/branches/ripleygmg_from_3668/ripley/src/Rectangle.cpp
ViewVC logotype

Annotation of /branches/ripleygmg_from_3668/ripley/src/Rectangle.cpp

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3748 - (hide annotations)
Thu Dec 15 07:36:19 2011 UTC (7 years, 11 months ago) by caltinay
File size: 115252 byte(s)
PDE assembly in serial and 2D seems to be doing what it's supposed to when
boundary elements are not involved.

1 caltinay 3691
2     /*******************************************************
3     *
4     * Copyright (c) 2003-2011 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     #include <ripley/Rectangle.h>
15     extern "C" {
16 caltinay 3748 #include "paso/SystemMatrix.h"
17 caltinay 3691 }
18    
19     #if USE_SILO
20     #include <silo.h>
21     #ifdef ESYS_MPI
22     #include <pmpio.h>
23     #endif
24     #endif
25    
26     #include <iomanip>
27    
28     using namespace std;
29    
30     namespace ripley {
31    
32     Rectangle::Rectangle(int n0, int n1, double l0, double l1, int d0, int d1) :
33     RipleyDomain(2),
34     m_gNE0(n0),
35     m_gNE1(n1),
36     m_l0(l0),
37     m_l1(l1),
38     m_NX(d0),
39     m_NY(d1)
40     {
41     // ensure number of subdivisions is valid and nodes can be distributed
42     // among number of ranks
43     if (m_NX*m_NY != m_mpiInfo->size)
44     throw RipleyException("Invalid number of spatial subdivisions");
45    
46     if (n0%m_NX > 0 || n1%m_NY > 0)
47     throw RipleyException("Number of elements must be separable into number of ranks in each dimension");
48    
49     // local number of elements
50     m_NE0 = n0/m_NX;
51     m_NE1 = n1/m_NY;
52     // local number of nodes (not necessarily owned)
53     m_N0 = m_NE0+1;
54     m_N1 = m_NE1+1;
55     // bottom-left node is at (offset0,offset1) in global mesh
56     m_offset0 = m_NE0*(m_mpiInfo->rank%m_NX);
57     m_offset1 = m_NE1*(m_mpiInfo->rank/m_NX);
58     populateSampleIds();
59     }
60    
61     Rectangle::~Rectangle()
62     {
63     }
64    
65     string Rectangle::getDescription() const
66     {
67     return "ripley::Rectangle";
68     }
69    
70     bool Rectangle::operator==(const AbstractDomain& other) const
71     {
72 caltinay 3744 const Rectangle* o=dynamic_cast<const Rectangle*>(&other);
73     if (o) {
74     return (RipleyDomain::operator==(other) &&
75     m_gNE0==o->m_gNE0 && m_gNE1==o->m_gNE1
76     && m_l0==o->m_l0 && m_l1==o->m_l1
77     && m_NX==o->m_NX && m_NY==o->m_NY);
78     }
79 caltinay 3691
80     return false;
81     }
82    
83     void Rectangle::dump(const string& fileName) const
84     {
85     #if USE_SILO
86     string fn(fileName);
87     if (fileName.length() < 6 || fileName.compare(fileName.length()-5, 5, ".silo") != 0) {
88     fn+=".silo";
89     }
90    
91     const int NUM_SILO_FILES = 1;
92     const char* blockDirFmt = "/block%04d";
93     int driver=DB_HDF5;
94     string siloPath;
95     DBfile* dbfile = NULL;
96    
97     #ifdef ESYS_MPI
98     PMPIO_baton_t* baton = NULL;
99     #endif
100    
101     if (m_mpiInfo->size > 1) {
102     #ifdef ESYS_MPI
103     baton = PMPIO_Init(NUM_SILO_FILES, PMPIO_WRITE, m_mpiInfo->comm,
104     0x1337, PMPIO_DefaultCreate, PMPIO_DefaultOpen,
105     PMPIO_DefaultClose, (void*)&driver);
106     // try the fallback driver in case of error
107     if (!baton && driver != DB_PDB) {
108     driver = DB_PDB;
109     baton = PMPIO_Init(NUM_SILO_FILES, PMPIO_WRITE, m_mpiInfo->comm,
110     0x1338, PMPIO_DefaultCreate, PMPIO_DefaultOpen,
111     PMPIO_DefaultClose, (void*)&driver);
112     }
113     if (baton) {
114     char str[64];
115     snprintf(str, 64, blockDirFmt, PMPIO_RankInGroup(baton, m_mpiInfo->rank));
116     siloPath = str;
117     dbfile = (DBfile*) PMPIO_WaitForBaton(baton, fn.c_str(), siloPath.c_str());
118     }
119     #endif
120     } else {
121     dbfile = DBCreate(fn.c_str(), DB_CLOBBER, DB_LOCAL,
122     getDescription().c_str(), driver);
123     // try the fallback driver in case of error
124     if (!dbfile && driver != DB_PDB) {
125     driver = DB_PDB;
126     dbfile = DBCreate(fn.c_str(), DB_CLOBBER, DB_LOCAL,
127     getDescription().c_str(), driver);
128     }
129     }
130    
131     if (!dbfile)
132     throw RipleyException("dump: Could not create Silo file");
133    
134     /*
135     if (driver==DB_HDF5) {
136     // gzip level 1 already provides good compression with minimal
137     // performance penalty. Some tests showed that gzip levels >3 performed
138     // rather badly on escript data both in terms of time and space
139     DBSetCompression("ERRMODE=FALLBACK METHOD=GZIP LEVEL=1");
140     }
141     */
142    
143     boost::scoped_ptr<double> x(new double[m_N0]);
144     boost::scoped_ptr<double> y(new double[m_N1]);
145     double* coords[2] = { x.get(), y.get() };
146 caltinay 3697 pair<double,double> xdx = getFirstCoordAndSpacing(0);
147     pair<double,double> ydy = getFirstCoordAndSpacing(1);
148 caltinay 3691 #pragma omp parallel
149     {
150 caltinay 3722 #pragma omp for nowait
151 caltinay 3691 for (dim_t i0 = 0; i0 < m_N0; i0++) {
152 caltinay 3697 coords[0][i0]=xdx.first+i0*xdx.second;
153 caltinay 3691 }
154 caltinay 3722 #pragma omp for nowait
155 caltinay 3691 for (dim_t i1 = 0; i1 < m_N1; i1++) {
156 caltinay 3697 coords[1][i1]=ydy.first+i1*ydy.second;
157 caltinay 3691 }
158     }
159 caltinay 3697 IndexVector dims = getNumNodesPerDim();
160    
161     // write mesh
162     DBPutQuadmesh(dbfile, "mesh", NULL, coords, &dims[0], 2, DB_DOUBLE,
163 caltinay 3691 DB_COLLINEAR, NULL);
164    
165 caltinay 3697 // write node ids
166     DBPutQuadvar1(dbfile, "nodeId", "mesh", (void*)&m_nodeId[0], &dims[0], 2,
167     NULL, 0, DB_INT, DB_NODECENT, NULL);
168    
169     // write element ids
170     dims = getNumElementsPerDim();
171     DBPutQuadvar1(dbfile, "elementId", "mesh", (void*)&m_elementId[0],
172     &dims[0], 2, NULL, 0, DB_INT, DB_ZONECENT, NULL);
173    
174     // rank 0 writes multimesh and multivar
175 caltinay 3691 if (m_mpiInfo->rank == 0) {
176     vector<string> tempstrings;
177 caltinay 3697 vector<char*> names;
178 caltinay 3691 for (dim_t i=0; i<m_mpiInfo->size; i++) {
179     stringstream path;
180     path << "/block" << setw(4) << setfill('0') << right << i << "/mesh";
181     tempstrings.push_back(path.str());
182 caltinay 3697 names.push_back((char*)tempstrings.back().c_str());
183 caltinay 3691 }
184 caltinay 3697 vector<int> types(m_mpiInfo->size, DB_QUAD_RECT);
185 caltinay 3691 DBSetDir(dbfile, "/");
186 caltinay 3697 DBPutMultimesh(dbfile, "multimesh", m_mpiInfo->size, &names[0],
187     &types[0], NULL);
188     tempstrings.clear();
189     names.clear();
190     for (dim_t i=0; i<m_mpiInfo->size; i++) {
191     stringstream path;
192     path << "/block" << setw(4) << setfill('0') << right << i << "/nodeId";
193     tempstrings.push_back(path.str());
194     names.push_back((char*)tempstrings.back().c_str());
195     }
196     types.assign(m_mpiInfo->size, DB_QUADVAR);
197     DBPutMultivar(dbfile, "nodeId", m_mpiInfo->size, &names[0],
198     &types[0], NULL);
199     tempstrings.clear();
200     names.clear();
201     for (dim_t i=0; i<m_mpiInfo->size; i++) {
202     stringstream path;
203     path << "/block" << setw(4) << setfill('0') << right << i << "/elementId";
204     tempstrings.push_back(path.str());
205     names.push_back((char*)tempstrings.back().c_str());
206     }
207     DBPutMultivar(dbfile, "elementId", m_mpiInfo->size, &names[0],
208     &types[0], NULL);
209 caltinay 3691 }
210    
211     if (m_mpiInfo->size > 1) {
212     #ifdef ESYS_MPI
213     PMPIO_HandOffBaton(baton, dbfile);
214     PMPIO_Finish(baton);
215     #endif
216     } else {
217     DBClose(dbfile);
218     }
219    
220     #else // USE_SILO
221     throw RipleyException("dump(): no Silo support");
222     #endif
223     }
224    
225 caltinay 3697 const int* Rectangle::borrowSampleReferenceIDs(int fsType) const
226 caltinay 3691 {
227 caltinay 3697 switch (fsType) {
228 caltinay 3691 case Nodes:
229 caltinay 3748 case ReducedNodes: //FIXME: reduced
230 caltinay 3691 return &m_nodeId[0];
231     case Elements:
232 caltinay 3733 case ReducedElements:
233 caltinay 3691 return &m_elementId[0];
234     case FaceElements:
235 caltinay 3733 case ReducedFaceElements:
236 caltinay 3691 return &m_faceId[0];
237     default:
238     break;
239     }
240    
241     stringstream msg;
242     msg << "borrowSampleReferenceIDs() not implemented for function space type "
243 caltinay 3701 << functionSpaceTypeAsString(fsType);
244 caltinay 3691 throw RipleyException(msg.str());
245     }
246    
247     bool Rectangle::ownSample(int fsCode, index_t id) const
248     {
249     #ifdef ESYS_MPI
250 caltinay 3702 if (fsCode != ReducedNodes) {
251 caltinay 3699 const index_t myFirst=m_nodeDistribution[m_mpiInfo->rank];
252     const index_t myLast=m_nodeDistribution[m_mpiInfo->rank+1]-1;
253 caltinay 3691 return (m_nodeId[id]>=myFirst && m_nodeId[id]<=myLast);
254 caltinay 3702 } else {
255     stringstream msg;
256     msg << "ownSample() not implemented for "
257     << functionSpaceTypeAsString(fsCode);
258     throw RipleyException(msg.str());
259     }
260 caltinay 3691 #else
261     return true;
262     #endif
263     }
264    
265 caltinay 3702 void Rectangle::setToGradient(escript::Data& out, const escript::Data& cIn) const
266 caltinay 3691 {
267 caltinay 3702 escript::Data& in = *const_cast<escript::Data*>(&cIn);
268 caltinay 3701 const dim_t numComp = in.getDataPointSize();
269 caltinay 3702 const double h0 = m_l0/m_gNE0;
270     const double h1 = m_l1/m_gNE1;
271 caltinay 3724 const double cx0 = -1./h0;
272     const double cx1 = -.78867513459481288225/h0;
273     const double cx2 = -.5/h0;
274     const double cx3 = -.21132486540518711775/h0;
275     const double cx4 = .21132486540518711775/h0;
276     const double cx5 = .5/h0;
277     const double cx6 = .78867513459481288225/h0;
278     const double cx7 = 1./h0;
279     const double cy0 = -1./h1;
280     const double cy1 = -.78867513459481288225/h1;
281     const double cy2 = -.5/h1;
282     const double cy3 = -.21132486540518711775/h1;
283     const double cy4 = .21132486540518711775/h1;
284     const double cy5 = .5/h1;
285     const double cy6 = .78867513459481288225/h1;
286     const double cy7 = 1./h1;
287    
288 caltinay 3702 if (out.getFunctionSpace().getTypeCode() == Elements) {
289 caltinay 3724 /*** GENERATOR SNIP_GRAD_ELEMENTS TOP */
290 caltinay 3697 #pragma omp parallel for
291 caltinay 3707 for (index_t k1=0; k1 < m_NE1; ++k1) {
292     for (index_t k0=0; k0 < m_NE0; ++k0) {
293 caltinay 3702 const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,k1, m_N0));
294     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,k1+1, m_N0));
295     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,k1+1, m_N0));
296     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,k1, m_N0));
297     double* o = out.getSampleDataRW(INDEX2(k0,k1,m_NE0));
298     for (index_t i=0; i < numComp; ++i) {
299 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_00[i]*cx1 + f_01[i]*cx3 + f_10[i]*cx6 + f_11[i]*cx4;
300     o[INDEX3(i,1,0,numComp,2)] = f_00[i]*cy1 + f_01[i]*cy6 + f_10[i]*cy3 + f_11[i]*cy4;
301     o[INDEX3(i,0,1,numComp,2)] = f_00[i]*cx1 + f_01[i]*cx3 + f_10[i]*cx6 + f_11[i]*cx4;
302     o[INDEX3(i,1,1,numComp,2)] = f_00[i]*cy3 + f_01[i]*cy4 + f_10[i]*cy1 + f_11[i]*cy6;
303     o[INDEX3(i,0,2,numComp,2)] = f_00[i]*cx3 + f_01[i]*cx1 + f_10[i]*cx4 + f_11[i]*cx6;
304     o[INDEX3(i,1,2,numComp,2)] = f_00[i]*cy1 + f_01[i]*cy6 + f_10[i]*cy3 + f_11[i]*cy4;
305     o[INDEX3(i,0,3,numComp,2)] = f_00[i]*cx3 + f_01[i]*cx1 + f_10[i]*cx4 + f_11[i]*cx6;
306     o[INDEX3(i,1,3,numComp,2)] = f_00[i]*cy3 + f_01[i]*cy4 + f_10[i]*cy1 + f_11[i]*cy6;
307 caltinay 3702 } /* end of component loop i */
308     } /* end of k0 loop */
309     } /* end of k1 loop */
310     /* GENERATOR SNIP_GRAD_ELEMENTS BOTTOM */
311 caltinay 3711 } else if (out.getFunctionSpace().getTypeCode() == ReducedElements) {
312 caltinay 3724 /*** GENERATOR SNIP_GRAD_REDUCED_ELEMENTS TOP */
313 caltinay 3711 #pragma omp parallel for
314     for (index_t k1=0; k1 < m_NE1; ++k1) {
315     for (index_t k0=0; k0 < m_NE0; ++k0) {
316     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,k1, m_N0));
317     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,k1+1, m_N0));
318     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,k1+1, m_N0));
319     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,k1, m_N0));
320     double* o = out.getSampleDataRW(INDEX2(k0,k1,m_NE0));
321     for (index_t i=0; i < numComp; ++i) {
322 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = cx5*(f_10[i] + f_11[i]) + cx2*(f_00[i] + f_01[i]);
323     o[INDEX3(i,1,0,numComp,2)] = cy2*(f_00[i] + f_10[i]) + cy5*(f_01[i] + f_11[i]);
324 caltinay 3711 } /* end of component loop i */
325     } /* end of k0 loop */
326     } /* end of k1 loop */
327     /* GENERATOR SNIP_GRAD_REDUCED_ELEMENTS BOTTOM */
328 caltinay 3707 } else if (out.getFunctionSpace().getTypeCode() == FaceElements) {
329 caltinay 3722 #pragma omp parallel
330     {
331 caltinay 3724 /*** GENERATOR SNIP_GRAD_FACES TOP */
332 caltinay 3722 if (m_faceOffset[0] > -1) {
333     #pragma omp for nowait
334     for (index_t k1=0; k1 < m_NE1; ++k1) {
335     const register double* f_10 = in.getSampleDataRO(INDEX2(1,k1, m_N0));
336     const register double* f_11 = in.getSampleDataRO(INDEX2(1,k1+1, m_N0));
337     const register double* f_01 = in.getSampleDataRO(INDEX2(0,k1+1, m_N0));
338     const register double* f_00 = in.getSampleDataRO(INDEX2(0,k1, m_N0));
339     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
340     for (index_t i=0; i < numComp; ++i) {
341 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_00[i]*cx1 + f_01[i]*cx3 + f_10[i]*cx6 + f_11[i]*cx4;
342     o[INDEX3(i,1,0,numComp,2)] = f_00[i]*cy0 + f_01[i]*cy7;
343     o[INDEX3(i,0,1,numComp,2)] = f_00[i]*cx3 + f_01[i]*cx1 + f_10[i]*cx4 + f_11[i]*cx6;
344     o[INDEX3(i,1,1,numComp,2)] = f_00[i]*cy0 + f_01[i]*cy7;
345 caltinay 3722 } /* end of component loop i */
346     } /* end of k1 loop */
347     } /* end of face 0 */
348     if (m_faceOffset[1] > -1) {
349     #pragma omp for nowait
350     for (index_t k1=0; k1 < m_NE1; ++k1) {
351     const register double* f_10 = in.getSampleDataRO(INDEX2(m_N0-1,k1, m_N0));
352     const register double* f_11 = in.getSampleDataRO(INDEX2(m_N0-1,k1+1, m_N0));
353     const register double* f_01 = in.getSampleDataRO(INDEX2(m_N0-2,k1+1, m_N0));
354     const register double* f_00 = in.getSampleDataRO(INDEX2(m_N0-2,k1, m_N0));
355     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
356     for (index_t i=0; i < numComp; ++i) {
357 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_00[i]*cx1 + f_01[i]*cx3 + f_10[i]*cx6 + f_11[i]*cx4;
358     o[INDEX3(i,1,0,numComp,2)] = f_10[i]*cy0 + f_11[i]*cy7;
359     o[INDEX3(i,0,1,numComp,2)] = f_00[i]*cx3 + f_01[i]*cx1 + f_10[i]*cx4 + f_11[i]*cx6;
360     o[INDEX3(i,1,1,numComp,2)] = f_10[i]*cy0 + f_11[i]*cy7;
361 caltinay 3722 } /* end of component loop i */
362     } /* end of k1 loop */
363     } /* end of face 1 */
364     if (m_faceOffset[2] > -1) {
365     #pragma omp for nowait
366     for (index_t k0=0; k0 < m_NE0; ++k0) {
367     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,0, m_N0));
368     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,0, m_N0));
369     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,1, m_N0));
370     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,1, m_N0));
371     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
372     for (index_t i=0; i < numComp; ++i) {
373 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_00[i]*cx0 + f_10[i]*cx7;
374     o[INDEX3(i,1,0,numComp,2)] = f_00[i]*cy1 + f_01[i]*cy6 + f_10[i]*cy3 + f_11[i]*cy4;
375     o[INDEX3(i,0,1,numComp,2)] = f_00[i]*cx0 + f_10[i]*cx7;
376     o[INDEX3(i,1,1,numComp,2)] = f_00[i]*cy3 + f_01[i]*cy4 + f_10[i]*cy1 + f_11[i]*cy6;
377 caltinay 3722 } /* end of component loop i */
378     } /* end of k0 loop */
379     } /* end of face 2 */
380     if (m_faceOffset[3] > -1) {
381     #pragma omp for nowait
382     for (index_t k0=0; k0 < m_NE0; ++k0) {
383     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,m_N1-1, m_N0));
384     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,m_N1-1, m_N0));
385     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,m_N1-2, m_N0));
386     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,m_N1-2, m_N0));
387     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
388     for (index_t i=0; i < numComp; ++i) {
389 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_01[i]*cx0 + f_11[i]*cx7;
390     o[INDEX3(i,1,0,numComp,2)] = f_00[i]*cy1 + f_01[i]*cy6 + f_10[i]*cy3 + f_11[i]*cy4;
391     o[INDEX3(i,0,1,numComp,2)] = f_01[i]*cx0 + f_11[i]*cx7;
392     o[INDEX3(i,1,1,numComp,2)] = f_00[i]*cy3 + f_01[i]*cy4 + f_10[i]*cy1 + f_11[i]*cy6;
393 caltinay 3722 } /* end of component loop i */
394     } /* end of k0 loop */
395     } /* end of face 3 */
396     /* GENERATOR SNIP_GRAD_FACES BOTTOM */
397     } // end of parallel section
398 caltinay 3711 } else if (out.getFunctionSpace().getTypeCode() == ReducedFaceElements) {
399 caltinay 3722 #pragma omp parallel
400     {
401 caltinay 3724 /*** GENERATOR SNIP_GRAD_REDUCED_FACES TOP */
402 caltinay 3722 if (m_faceOffset[0] > -1) {
403     #pragma omp for nowait
404     for (index_t k1=0; k1 < m_NE1; ++k1) {
405     const register double* f_10 = in.getSampleDataRO(INDEX2(1,k1, m_N0));
406     const register double* f_11 = in.getSampleDataRO(INDEX2(1,k1+1, m_N0));
407     const register double* f_01 = in.getSampleDataRO(INDEX2(0,k1+1, m_N0));
408     const register double* f_00 = in.getSampleDataRO(INDEX2(0,k1, m_N0));
409     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
410     for (index_t i=0; i < numComp; ++i) {
411 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = cx5*(f_10[i] + f_11[i]) + cx2*(f_00[i] + f_01[i]);
412     o[INDEX3(i,1,0,numComp,2)] = f_00[i]*cy0 + f_01[i]*cy7;
413 caltinay 3722 } /* end of component loop i */
414     } /* end of k1 loop */
415     } /* end of face 0 */
416     if (m_faceOffset[1] > -1) {
417     #pragma omp for nowait
418     for (index_t k1=0; k1 < m_NE1; ++k1) {
419     const register double* f_10 = in.getSampleDataRO(INDEX2(m_N0-1,k1, m_N0));
420     const register double* f_11 = in.getSampleDataRO(INDEX2(m_N0-1,k1+1, m_N0));
421     const register double* f_01 = in.getSampleDataRO(INDEX2(m_N0-2,k1+1, m_N0));
422     const register double* f_00 = in.getSampleDataRO(INDEX2(m_N0-2,k1, m_N0));
423     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
424     for (index_t i=0; i < numComp; ++i) {
425 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = cx5*(f_10[i] + f_11[i]) + cx2*(f_00[i] + f_01[i]);
426     o[INDEX3(i,1,0,numComp,2)] = f_10[i]*cy0 + f_11[i]*cy7;
427 caltinay 3722 } /* end of component loop i */
428     } /* end of k1 loop */
429     } /* end of face 1 */
430     if (m_faceOffset[2] > -1) {
431     #pragma omp for nowait
432     for (index_t k0=0; k0 < m_NE0; ++k0) {
433     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,0, m_N0));
434     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,0, m_N0));
435     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,1, m_N0));
436     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,1, m_N0));
437     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
438     for (index_t i=0; i < numComp; ++i) {
439 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_00[i]*cx0 + f_10[i]*cx7;
440     o[INDEX3(i,1,0,numComp,2)] = cy2*(f_00[i] + f_10[i]) + cy5*(f_01[i] + f_11[i]);
441 caltinay 3722 } /* end of component loop i */
442     } /* end of k0 loop */
443     } /* end of face 2 */
444     if (m_faceOffset[3] > -1) {
445     #pragma omp for nowait
446     for (index_t k0=0; k0 < m_NE0; ++k0) {
447     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,m_N1-1, m_N0));
448     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,m_N1-1, m_N0));
449     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,m_N1-2, m_N0));
450     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,m_N1-2, m_N0));
451     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
452     for (index_t i=0; i < numComp; ++i) {
453 caltinay 3724 o[INDEX3(i,0,0,numComp,2)] = f_01[i]*cx0 + f_11[i]*cx7;
454     o[INDEX3(i,1,0,numComp,2)] = cy5*(f_01[i] + f_11[i]) + cy2*(f_00[i] + f_10[i]);
455 caltinay 3722 } /* end of component loop i */
456     } /* end of k0 loop */
457     } /* end of face 3 */
458     /* GENERATOR SNIP_GRAD_REDUCED_FACES BOTTOM */
459     } // end of parallel section
460 caltinay 3702 } else {
461 caltinay 3707 stringstream msg;
462     msg << "setToGradient() not implemented for "
463     << functionSpaceTypeAsString(out.getFunctionSpace().getTypeCode());
464     throw RipleyException(msg.str());
465 caltinay 3702 }
466 caltinay 3701 }
467 caltinay 3697
468 caltinay 3713 void Rectangle::setToIntegrals(vector<double>& integrals, const escript::Data& arg) const
469     {
470     escript::Data& in = *const_cast<escript::Data*>(&arg);
471     const dim_t numComp = in.getDataPointSize();
472     const double h0 = m_l0/m_gNE0;
473     const double h1 = m_l1/m_gNE1;
474     if (arg.getFunctionSpace().getTypeCode() == Elements) {
475     const double w_0 = h0*h1/4.;
476     #pragma omp parallel
477     {
478     vector<double> int_local(numComp, 0);
479 caltinay 3722 #pragma omp for nowait
480 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
481     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
482     const double* f = in.getSampleDataRO(INDEX2(k0, k1, m_NE0));
483     for (index_t i=0; i < numComp; ++i) {
484     const register double f_0 = f[INDEX2(i,0,numComp)];
485     const register double f_1 = f[INDEX2(i,1,numComp)];
486     const register double f_2 = f[INDEX2(i,2,numComp)];
487     const register double f_3 = f[INDEX2(i,3,numComp)];
488     int_local[i]+=(f_0+f_1+f_2+f_3)*w_0;
489     } /* end of component loop i */
490     } /* end of k0 loop */
491     } /* end of k1 loop */
492    
493     #pragma omp critical
494     for (index_t i=0; i<numComp; i++)
495     integrals[i]+=int_local[i];
496 caltinay 3722 } // end of parallel section
497 caltinay 3713 } else if (arg.getFunctionSpace().getTypeCode() == ReducedElements) {
498     const double w_0 = h0*h1;
499     #pragma omp parallel
500     {
501     vector<double> int_local(numComp, 0);
502 caltinay 3722 #pragma omp for nowait
503 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
504     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
505     const double* f = in.getSampleDataRO(INDEX2(k0, k1, m_NE0));
506     for (index_t i=0; i < numComp; ++i) {
507     int_local[i]+=f[i]*w_0;
508     } /* end of component loop i */
509     } /* end of k0 loop */
510     } /* end of k1 loop */
511    
512     #pragma omp critical
513     for (index_t i=0; i<numComp; i++)
514     integrals[i]+=int_local[i];
515 caltinay 3722 } // end of parallel section
516 caltinay 3713 } else if (arg.getFunctionSpace().getTypeCode() == FaceElements) {
517     const double w_0 = h0/2.;
518     const double w_1 = h1/2.;
519     #pragma omp parallel
520     {
521     vector<double> int_local(numComp, 0);
522     if (m_faceOffset[0] > -1) {
523 caltinay 3722 #pragma omp for nowait
524 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
525     const double* f = in.getSampleDataRO(m_faceOffset[0]+k1);
526     for (index_t i=0; i < numComp; ++i) {
527     const register double f_0 = f[INDEX2(i,0,numComp)];
528     const register double f_1 = f[INDEX2(i,1,numComp)];
529     int_local[i]+=(f_0+f_1)*w_1;
530     } /* end of component loop i */
531     } /* end of k1 loop */
532     }
533    
534     if (m_faceOffset[1] > -1) {
535 caltinay 3722 #pragma omp for nowait
536 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
537     const double* f = in.getSampleDataRO(m_faceOffset[1]+k1);
538     for (index_t i=0; i < numComp; ++i) {
539     const register double f_0 = f[INDEX2(i,0,numComp)];
540     const register double f_1 = f[INDEX2(i,1,numComp)];
541     int_local[i]+=(f_0+f_1)*w_1;
542     } /* end of component loop i */
543     } /* end of k1 loop */
544     }
545    
546     if (m_faceOffset[2] > -1) {
547 caltinay 3722 #pragma omp for nowait
548 caltinay 3713 for (index_t k0 = 0; k0 < m_NE0; ++k0) {
549     const double* f = in.getSampleDataRO(m_faceOffset[2]+k0);
550     for (index_t i=0; i < numComp; ++i) {
551     const register double f_0 = f[INDEX2(i,0,numComp)];
552     const register double f_1 = f[INDEX2(i,1,numComp)];
553     int_local[i]+=(f_0+f_1)*w_0;
554     } /* end of component loop i */
555     } /* end of k0 loop */
556     }
557    
558     if (m_faceOffset[3] > -1) {
559 caltinay 3722 #pragma omp for nowait
560 caltinay 3713 for (index_t k0 = 0; k0 < m_NE0; ++k0) {
561     const double* f = in.getSampleDataRO(m_faceOffset[3]+k0);
562     for (index_t i=0; i < numComp; ++i) {
563     const register double f_0 = f[INDEX2(i,0,numComp)];
564     const register double f_1 = f[INDEX2(i,1,numComp)];
565     int_local[i]+=(f_0+f_1)*w_0;
566     } /* end of component loop i */
567     } /* end of k0 loop */
568     }
569    
570     #pragma omp critical
571     for (index_t i=0; i<numComp; i++)
572     integrals[i]+=int_local[i];
573 caltinay 3722 } // end of parallel section
574 caltinay 3713 } else if (arg.getFunctionSpace().getTypeCode() == ReducedFaceElements) {
575     #pragma omp parallel
576     {
577     vector<double> int_local(numComp, 0);
578     if (m_faceOffset[0] > -1) {
579 caltinay 3722 #pragma omp for nowait
580 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
581     const double* f = in.getSampleDataRO(m_faceOffset[0]+k1);
582     for (index_t i=0; i < numComp; ++i) {
583     int_local[i]+=f[i]*h1;
584     } /* end of component loop i */
585     } /* end of k1 loop */
586     }
587    
588     if (m_faceOffset[1] > -1) {
589 caltinay 3722 #pragma omp for nowait
590 caltinay 3713 for (index_t k1 = 0; k1 < m_NE1; ++k1) {
591     const double* f = in.getSampleDataRO(m_faceOffset[1]+k1);
592     for (index_t i=0; i < numComp; ++i) {
593     int_local[i]+=f[i]*h1;
594     } /* end of component loop i */
595     } /* end of k1 loop */
596     }
597    
598     if (m_faceOffset[2] > -1) {
599 caltinay 3722 #pragma omp for nowait
600 caltinay 3713 for (index_t k0 = 0; k0 < m_NE0; ++k0) {
601     const double* f = in.getSampleDataRO(m_faceOffset[2]+k0);
602     for (index_t i=0; i < numComp; ++i) {
603     int_local[i]+=f[i]*h0;
604     } /* end of component loop i */
605     } /* end of k0 loop */
606     }
607    
608     if (m_faceOffset[3] > -1) {
609 caltinay 3722 #pragma omp for nowait
610 caltinay 3713 for (index_t k0 = 0; k0 < m_NE0; ++k0) {
611     const double* f = in.getSampleDataRO(m_faceOffset[3]+k0);
612     for (index_t i=0; i < numComp; ++i) {
613     int_local[i]+=f[i]*h0;
614     } /* end of component loop i */
615     } /* end of k0 loop */
616     }
617    
618     #pragma omp critical
619     for (index_t i=0; i<numComp; i++)
620     integrals[i]+=int_local[i];
621 caltinay 3722 } // end of parallel section
622 caltinay 3713 } else {
623     stringstream msg;
624     msg << "setToIntegrals() not implemented for "
625     << functionSpaceTypeAsString(arg.getFunctionSpace().getTypeCode());
626     throw RipleyException(msg.str());
627     }
628     }
629    
630 caltinay 3722 void Rectangle::setToNormal(escript::Data& out) const
631     {
632     if (out.getFunctionSpace().getTypeCode() == FaceElements) {
633     #pragma omp parallel
634     {
635     if (m_faceOffset[0] > -1) {
636     #pragma omp for nowait
637     for (index_t k1 = 0; k1 < m_NE1; ++k1) {
638     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
639     // set vector at two quadrature points
640     *o++ = -1.;
641     *o++ = 0.;
642     *o++ = -1.;
643     *o = 0.;
644     }
645     }
646    
647     if (m_faceOffset[1] > -1) {
648     #pragma omp for nowait
649     for (index_t k1 = 0; k1 < m_NE1; ++k1) {
650     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
651     // set vector at two quadrature points
652     *o++ = 1.;
653     *o++ = 0.;
654     *o++ = 1.;
655     *o = 0.;
656     }
657     }
658    
659     if (m_faceOffset[2] > -1) {
660     #pragma omp for nowait
661     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
662     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
663     // set vector at two quadrature points
664     *o++ = 0.;
665     *o++ = -1.;
666     *o++ = 0.;
667     *o = -1.;
668     }
669     }
670    
671     if (m_faceOffset[3] > -1) {
672     #pragma omp for nowait
673     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
674     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
675     // set vector at two quadrature points
676     *o++ = 0.;
677     *o++ = 1.;
678     *o++ = 0.;
679     *o = 1.;
680     }
681     }
682     } // end of parallel section
683     } else if (out.getFunctionSpace().getTypeCode() == ReducedFaceElements) {
684     #pragma omp parallel
685     {
686     if (m_faceOffset[0] > -1) {
687     #pragma omp for nowait
688     for (index_t k1 = 0; k1 < m_NE1; ++k1) {
689     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
690     *o++ = -1.;
691     *o = 0.;
692     }
693     }
694    
695     if (m_faceOffset[1] > -1) {
696     #pragma omp for nowait
697     for (index_t k1 = 0; k1 < m_NE1; ++k1) {
698     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
699     *o++ = 1.;
700     *o = 0.;
701     }
702     }
703    
704     if (m_faceOffset[2] > -1) {
705     #pragma omp for nowait
706     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
707     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
708     *o++ = 0.;
709     *o = -1.;
710     }
711     }
712    
713     if (m_faceOffset[3] > -1) {
714     #pragma omp for nowait
715     for (index_t k0 = 0; k0 < m_NE0; ++k0) {
716     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
717     *o++ = 0.;
718     *o = 1.;
719     }
720     }
721     } // end of parallel section
722    
723     } else {
724     stringstream msg;
725     msg << "setToNormal() not implemented for "
726     << functionSpaceTypeAsString(out.getFunctionSpace().getTypeCode());
727     throw RipleyException(msg.str());
728     }
729     }
730    
731 caltinay 3691 Paso_SystemMatrixPattern* Rectangle::getPattern(bool reducedRowOrder,
732     bool reducedColOrder) const
733     {
734     if (reducedRowOrder || reducedColOrder)
735     throw RipleyException("getPattern() not implemented for reduced order");
736    
737 caltinay 3699 // connector
738     RankVector neighbour;
739     IndexVector offsetInShared(1,0);
740     IndexVector sendShared, recvShared;
741     const IndexVector faces=getNumFacesPerBoundary();
742     const index_t left = (m_offset0==0 ? 0 : 1);
743     const index_t bottom = (m_offset1==0 ? 0 : 1);
744     // corner node from bottom-left
745     if (faces[0] == 0 && faces[2] == 0) {
746     neighbour.push_back(m_mpiInfo->rank-m_NX-1);
747     offsetInShared.push_back(offsetInShared.back()+1);
748     sendShared.push_back(m_nodeId[m_N0+left]);
749     recvShared.push_back(m_nodeId[0]);
750     }
751     // bottom edge
752     if (faces[2] == 0) {
753     neighbour.push_back(m_mpiInfo->rank-m_NX);
754     offsetInShared.push_back(offsetInShared.back()+m_N0-left);
755     for (dim_t i=left; i<m_N0; i++) {
756     // easy case, we know the neighbour id's
757     sendShared.push_back(m_nodeId[m_N0+i]);
758     recvShared.push_back(m_nodeId[i]);
759     }
760     }
761     // corner node from bottom-right
762     if (faces[1] == 0 && faces[2] == 0) {
763     neighbour.push_back(m_mpiInfo->rank-m_NX+1);
764     const index_t N0=(neighbour.back()%m_NX == 0 ? m_N0 : m_N0-1);
765     const index_t N1=(neighbour.back()/m_NX == 0 ? m_N1 : m_N1-1);
766     const index_t first=m_nodeDistribution[neighbour.back()];
767     offsetInShared.push_back(offsetInShared.back()+1);
768     sendShared.push_back(m_nodeId[(bottom+1)*m_N0-1]);
769     recvShared.push_back(first+N0*(N1-1));
770     }
771     // left edge
772     if (faces[0] == 0) {
773     neighbour.push_back(m_mpiInfo->rank-1);
774     offsetInShared.push_back(offsetInShared.back()+m_N1-bottom);
775     for (dim_t i=bottom; i<m_N1; i++) {
776     // easy case, we know the neighbour id's
777     sendShared.push_back(m_nodeId[i*m_N0+1]);
778     recvShared.push_back(m_nodeId[i*m_N0]);
779     }
780     }
781     // right edge
782     if (faces[1] == 0) {
783     neighbour.push_back(m_mpiInfo->rank+1);
784     const index_t rightN0=(neighbour.back()%m_NX == 0 ? m_N0 : m_N0-1);
785     const index_t first=m_nodeDistribution[neighbour.back()];
786     offsetInShared.push_back(offsetInShared.back()+m_N1-bottom);
787     for (dim_t i=bottom; i<m_N1; i++) {
788     sendShared.push_back(m_nodeId[(i+1)*m_N0-1]);
789     recvShared.push_back(first+rightN0*(i-bottom));
790     }
791     }
792     // corner node from top-left
793     if (faces[0] == 0 && faces[3] == 0) {
794     neighbour.push_back(m_mpiInfo->rank+m_NX-1);
795     const index_t N0=(neighbour.back()%m_NX == 0 ? m_N0 : m_N0-1);
796     const index_t first=m_nodeDistribution[neighbour.back()];
797     offsetInShared.push_back(offsetInShared.back()+1);
798     sendShared.push_back(m_nodeId[m_N0*(m_N1-1)+left]);
799     recvShared.push_back(first+N0-1);
800     }
801     // top edge
802     if (faces[3] == 0) {
803     neighbour.push_back(m_mpiInfo->rank+m_NX);
804     const index_t first=m_nodeDistribution[neighbour.back()];
805     offsetInShared.push_back(offsetInShared.back()+m_N0-left);
806     for (dim_t i=left; i<m_N0; i++) {
807     sendShared.push_back(m_nodeId[m_N0*(m_N1-1)+i]);
808     recvShared.push_back(first+i-left);
809     }
810     }
811     // corner node from top-right
812     if (faces[1] == 0 && faces[3] == 0) {
813     neighbour.push_back(m_mpiInfo->rank+m_NX+1);
814     const index_t first=m_nodeDistribution[neighbour.back()];
815     offsetInShared.push_back(offsetInShared.back()+1);
816     sendShared.push_back(m_nodeId[m_N0*m_N1-1]);
817     recvShared.push_back(first);
818     }
819     const int numDOF=m_nodeDistribution[m_mpiInfo->rank+1]-m_nodeDistribution[m_mpiInfo->rank];
820 caltinay 3702 /*
821 caltinay 3699 cout << "--- rcv_shcomp ---" << endl;
822     cout << "numDOF=" << numDOF << ", numNeighbors=" << neighbour.size() << endl;
823     for (size_t i=0; i<neighbour.size(); i++) {
824     cout << "neighbor[" << i << "]=" << neighbour[i]
825     << " offsetInShared[" << i+1 << "]=" << offsetInShared[i+1] << endl;
826     }
827     for (size_t i=0; i<recvShared.size(); i++) {
828     cout << "shared[" << i << "]=" << recvShared[i] << endl;
829     }
830     cout << "--- snd_shcomp ---" << endl;
831     for (size_t i=0; i<sendShared.size(); i++) {
832     cout << "shared[" << i << "]=" << sendShared[i] << endl;
833     }
834 caltinay 3702 */
835 caltinay 3691
836     Paso_SharedComponents *snd_shcomp = Paso_SharedComponents_alloc(
837 caltinay 3699 numDOF, neighbour.size(), &neighbour[0], &sendShared[0],
838 caltinay 3697 &offsetInShared[0], 1, 0, m_mpiInfo);
839 caltinay 3691 Paso_SharedComponents *rcv_shcomp = Paso_SharedComponents_alloc(
840 caltinay 3699 numDOF, neighbour.size(), &neighbour[0], &recvShared[0],
841 caltinay 3697 &offsetInShared[0], 1, 0, m_mpiInfo);
842 caltinay 3691 Paso_Connector* connector = Paso_Connector_alloc(snd_shcomp, rcv_shcomp);
843 caltinay 3699 Paso_SharedComponents_free(snd_shcomp);
844     Paso_SharedComponents_free(rcv_shcomp);
845 caltinay 3691
846     // create patterns
847 caltinay 3699 dim_t M, N;
848     IndexVector ptr(1,0);
849     IndexVector index;
850    
851     // main pattern
852     for (index_t i=0; i<numDOF; i++) {
853     // always add the node itself
854     index.push_back(i);
855 caltinay 3748 const int num=insertNeighbours(index, i);
856 caltinay 3699 ptr.push_back(ptr.back()+num+1);
857     }
858     M=N=ptr.size()-1;
859     // paso will manage the memory
860     index_t* indexC = MEMALLOC(index.size(),index_t);
861     index_t* ptrC = MEMALLOC(ptr.size(), index_t);
862     copy(index.begin(), index.end(), indexC);
863     copy(ptr.begin(), ptr.end(), ptrC);
864 caltinay 3691 Paso_Pattern *mainPattern = Paso_Pattern_alloc(MATRIX_FORMAT_DEFAULT,
865 caltinay 3699 M, N, ptrC, indexC);
866 caltinay 3691
867 caltinay 3702 /*
868 caltinay 3699 cout << "--- main_pattern ---" << endl;
869     cout << "M=" << M << ", N=" << N << endl;
870     for (size_t i=0; i<ptr.size(); i++) {
871     cout << "ptr[" << i << "]=" << ptr[i] << endl;
872     }
873     for (size_t i=0; i<index.size(); i++) {
874     cout << "index[" << i << "]=" << index[i] << endl;
875     }
876 caltinay 3702 */
877 caltinay 3699
878     ptr.clear();
879     index.clear();
880    
881     // column & row couple patterns
882     Paso_Pattern *colCouplePattern, *rowCouplePattern;
883     generateCouplePatterns(&colCouplePattern, &rowCouplePattern);
884    
885     // allocate paso distribution
886     Paso_Distribution* distribution = Paso_Distribution_alloc(m_mpiInfo,
887     const_cast<index_t*>(&m_nodeDistribution[0]), 1, 0);
888    
889 caltinay 3691 Paso_SystemMatrixPattern* pattern = Paso_SystemMatrixPattern_alloc(
890     MATRIX_FORMAT_DEFAULT, distribution, distribution,
891     mainPattern, colCouplePattern, rowCouplePattern,
892     connector, connector);
893     Paso_Pattern_free(mainPattern);
894     Paso_Pattern_free(colCouplePattern);
895     Paso_Pattern_free(rowCouplePattern);
896     Paso_Distribution_free(distribution);
897 caltinay 3697 return pattern;
898 caltinay 3691 }
899    
900     void Rectangle::Print_Mesh_Info(const bool full) const
901     {
902     RipleyDomain::Print_Mesh_Info(full);
903     if (full) {
904     cout << " Id Coordinates" << endl;
905     cout.precision(15);
906     cout.setf(ios::scientific, ios::floatfield);
907 caltinay 3697 pair<double,double> xdx = getFirstCoordAndSpacing(0);
908     pair<double,double> ydy = getFirstCoordAndSpacing(1);
909 caltinay 3691 for (index_t i=0; i < getNumNodes(); i++) {
910     cout << " " << setw(5) << m_nodeId[i]
911 caltinay 3697 << " " << xdx.first+(i%m_N0)*xdx.second
912     << " " << ydy.first+(i/m_N0)*ydy.second << endl;
913 caltinay 3691 }
914     }
915     }
916    
917 caltinay 3697 IndexVector Rectangle::getNumNodesPerDim() const
918     {
919     IndexVector ret;
920     ret.push_back(m_N0);
921     ret.push_back(m_N1);
922     return ret;
923     }
924    
925     IndexVector Rectangle::getNumElementsPerDim() const
926     {
927     IndexVector ret;
928     ret.push_back(m_NE0);
929     ret.push_back(m_NE1);
930     return ret;
931     }
932    
933     IndexVector Rectangle::getNumFacesPerBoundary() const
934     {
935     IndexVector ret(4, 0);
936     //left
937     if (m_offset0==0)
938     ret[0]=m_NE1;
939     //right
940     if (m_mpiInfo->rank%m_NX==m_NX-1)
941     ret[1]=m_NE1;
942     //bottom
943     if (m_offset1==0)
944     ret[2]=m_NE0;
945     //top
946     if (m_mpiInfo->rank/m_NX==m_NY-1)
947     ret[3]=m_NE0;
948     return ret;
949     }
950    
951     pair<double,double> Rectangle::getFirstCoordAndSpacing(dim_t dim) const
952     {
953     if (dim==0) {
954     return pair<double,double>((m_l0*m_offset0)/m_gNE0, m_l0/m_gNE0);
955     } else if (dim==1) {
956     return pair<double,double>((m_l1*m_offset1)/m_gNE1, m_l1/m_gNE1);
957     }
958     throw RipleyException("getFirstCoordAndSpacing(): invalid argument");
959     }
960    
961 caltinay 3691 //protected
962     dim_t Rectangle::getNumFaceElements() const
963     {
964 caltinay 3699 const IndexVector faces = getNumFacesPerBoundary();
965 caltinay 3691 dim_t n=0;
966 caltinay 3699 for (size_t i=0; i<faces.size(); i++)
967     n+=faces[i];
968 caltinay 3691 return n;
969     }
970    
971     //protected
972     void Rectangle::assembleCoordinates(escript::Data& arg) const
973     {
974     escriptDataC x = arg.getDataC();
975     int numDim = m_numDim;
976     if (!isDataPointShapeEqual(&x, 1, &numDim))
977     throw RipleyException("setToX: Invalid Data object shape");
978     if (!numSamplesEqual(&x, 1, getNumNodes()))
979     throw RipleyException("setToX: Illegal number of samples in Data object");
980    
981 caltinay 3697 pair<double,double> xdx = getFirstCoordAndSpacing(0);
982     pair<double,double> ydy = getFirstCoordAndSpacing(1);
983 caltinay 3691 arg.requireWrite();
984     #pragma omp parallel for
985     for (dim_t i1 = 0; i1 < m_N1; i1++) {
986     for (dim_t i0 = 0; i0 < m_N0; i0++) {
987     double* point = arg.getSampleDataRW(i0+m_N0*i1);
988 caltinay 3697 point[0] = xdx.first+i0*xdx.second;
989     point[1] = ydy.first+i1*ydy.second;
990 caltinay 3691 }
991     }
992     }
993    
994     //private
995     void Rectangle::populateSampleIds()
996     {
997 caltinay 3697 // identifiers are ordered from left to right, bottom to top on each rank,
998     // except for the shared nodes which are owned by the rank below / to the
999     // left of the current rank
1000    
1001     // build node distribution vector first.
1002     // m_nodeDistribution[i] is the first node id on rank i, that is
1003     // rank i owns m_nodeDistribution[i+1]-nodeDistribution[i] nodes
1004     m_nodeDistribution.assign(m_mpiInfo->size+1, 0);
1005     m_nodeDistribution[1]=getNumNodes();
1006     for (dim_t k=1; k<m_mpiInfo->size-1; k++) {
1007     const index_t x=k%m_NX;
1008     const index_t y=k/m_NX;
1009     index_t numNodes=getNumNodes();
1010     if (x>0)
1011     numNodes-=m_N1;
1012     if (y>0)
1013     numNodes-=m_N0;
1014     if (x>0 && y>0)
1015     numNodes++; // subtracted corner twice -> fix that
1016     m_nodeDistribution[k+1]=m_nodeDistribution[k]+numNodes;
1017     }
1018     m_nodeDistribution[m_mpiInfo->size]=getNumDataPointsGlobal();
1019    
1020 caltinay 3691 m_nodeId.resize(getNumNodes());
1021 caltinay 3697
1022     // the bottom row and left column are not owned by this rank so the
1023     // identifiers need to be computed accordingly
1024     const index_t left = (m_offset0==0 ? 0 : 1);
1025     const index_t bottom = (m_offset1==0 ? 0 : 1);
1026     if (left>0) {
1027     const int neighbour=m_mpiInfo->rank-1;
1028     const index_t leftN0=(neighbour%m_NX == 0 ? m_N0 : m_N0-1);
1029 caltinay 3691 #pragma omp parallel for
1030 caltinay 3697 for (dim_t i1=bottom; i1<m_N1; i1++) {
1031     m_nodeId[i1*m_N0]=m_nodeDistribution[neighbour]
1032     + (i1-bottom+1)*leftN0-1;
1033     }
1034     }
1035     if (bottom>0) {
1036     const int neighbour=m_mpiInfo->rank-m_NX;
1037     const index_t bottomN0=(neighbour%m_NX == 0 ? m_N0 : m_N0-1);
1038     const index_t bottomN1=(neighbour/m_NX == 0 ? m_N1 : m_N1-1);
1039     #pragma omp parallel for
1040     for (dim_t i0=left; i0<m_N0; i0++) {
1041     m_nodeId[i0]=m_nodeDistribution[neighbour]
1042     + (bottomN1-1)*bottomN0 + i0 - left;
1043     }
1044     }
1045     if (left>0 && bottom>0) {
1046     const int neighbour=m_mpiInfo->rank-m_NX-1;
1047 caltinay 3699 const index_t N0=(neighbour%m_NX == 0 ? m_N0 : m_N0-1);
1048     const index_t N1=(neighbour/m_NX == 0 ? m_N1 : m_N1-1);
1049     m_nodeId[0]=m_nodeDistribution[neighbour]+N0*N1-1;
1050 caltinay 3697 }
1051    
1052     // the rest of the id's are contiguous
1053     const index_t firstId=m_nodeDistribution[m_mpiInfo->rank];
1054     #pragma omp parallel for
1055     for (dim_t i1=bottom; i1<m_N1; i1++) {
1056     for (dim_t i0=left; i0<m_N0; i0++) {
1057 caltinay 3746 m_nodeId[i0+i1*m_N0] = firstId+i0-left+(i1-bottom)*(m_N0-left);
1058 caltinay 3697 }
1059     }
1060 caltinay 3735 m_nodeTags.assign(getNumNodes(), 0);
1061     updateTagsInUse(Nodes);
1062 caltinay 3697
1063 caltinay 3735 // elements
1064 caltinay 3697 m_elementId.resize(getNumElements());
1065     #pragma omp parallel for
1066     for (dim_t k=0; k<getNumElements(); k++) {
1067     m_elementId[k]=k;
1068     }
1069 caltinay 3735 m_elementTags.assign(getNumElements(), 0);
1070     updateTagsInUse(Elements);
1071 caltinay 3697
1072 caltinay 3704 // face element id's
1073 caltinay 3697 m_faceId.resize(getNumFaceElements());
1074     #pragma omp parallel for
1075     for (dim_t k=0; k<getNumFaceElements(); k++) {
1076     m_faceId[k]=k;
1077     }
1078 caltinay 3704
1079 caltinay 3722 // generate face offset vector and set face tags
1080 caltinay 3704 const IndexVector facesPerEdge = getNumFacesPerBoundary();
1081 caltinay 3722 const index_t LEFT=1, RIGHT=2, BOTTOM=10, TOP=20;
1082     const index_t faceTag[] = { LEFT, RIGHT, BOTTOM, TOP };
1083 caltinay 3704 m_faceOffset.assign(facesPerEdge.size(), -1);
1084 caltinay 3722 m_faceTags.clear();
1085 caltinay 3704 index_t offset=0;
1086     for (size_t i=0; i<facesPerEdge.size(); i++) {
1087     if (facesPerEdge[i]>0) {
1088     m_faceOffset[i]=offset;
1089     offset+=facesPerEdge[i];
1090 caltinay 3722 m_faceTags.insert(m_faceTags.end(), facesPerEdge[i], faceTag[i]);
1091 caltinay 3704 }
1092     }
1093 caltinay 3722 setTagMap("left", LEFT);
1094     setTagMap("right", RIGHT);
1095     setTagMap("bottom", BOTTOM);
1096     setTagMap("top", TOP);
1097     updateTagsInUse(FaceElements);
1098 caltinay 3691 }
1099    
1100 caltinay 3699 //private
1101     int Rectangle::insertNeighbours(IndexVector& index, index_t node) const
1102     {
1103     const dim_t myN0 = (m_offset0==0 ? m_N0 : m_N0-1);
1104     const dim_t myN1 = (m_offset1==0 ? m_N1 : m_N1-1);
1105     const int x=node%myN0;
1106     const int y=node/myN0;
1107     int num=0;
1108     if (y>0) {
1109     if (x>0) {
1110     // bottom-left
1111     index.push_back(node-myN0-1);
1112     num++;
1113     }
1114     // bottom
1115     index.push_back(node-myN0);
1116     num++;
1117     if (x<myN0-1) {
1118     // bottom-right
1119     index.push_back(node-myN0+1);
1120     num++;
1121     }
1122     }
1123     if (x<myN0-1) {
1124     // right
1125     index.push_back(node+1);
1126     num++;
1127     if (y<myN1-1) {
1128     // top-right
1129     index.push_back(node+myN0+1);
1130     num++;
1131     }
1132     }
1133     if (y<myN1-1) {
1134     // top
1135     index.push_back(node+myN0);
1136     num++;
1137     if (x>0) {
1138     // top-left
1139     index.push_back(node+myN0-1);
1140     num++;
1141     }
1142     }
1143     if (x>0) {
1144     // left
1145     index.push_back(node-1);
1146     num++;
1147     }
1148    
1149     return num;
1150     }
1151    
1152     //private
1153     void Rectangle::generateCouplePatterns(Paso_Pattern** colPattern, Paso_Pattern** rowPattern) const
1154     {
1155     IndexVector ptr(1,0);
1156     IndexVector index;
1157     const dim_t myN0 = (m_offset0==0 ? m_N0 : m_N0-1);
1158     const dim_t myN1 = (m_offset1==0 ? m_N1 : m_N1-1);
1159     const IndexVector faces=getNumFacesPerBoundary();
1160    
1161     // bottom edge
1162     dim_t n=0;
1163     if (faces[0] == 0) {
1164     index.push_back(2*(myN0+myN1+1));
1165     index.push_back(2*(myN0+myN1+1)+1);
1166     n+=2;
1167     if (faces[2] == 0) {
1168     index.push_back(0);
1169     index.push_back(1);
1170     index.push_back(2);
1171     n+=3;
1172     }
1173     } else if (faces[2] == 0) {
1174     index.push_back(1);
1175     index.push_back(2);
1176     n+=2;
1177     }
1178     // n=neighbours of bottom-left corner node
1179     ptr.push_back(ptr.back()+n);
1180     n=0;
1181     if (faces[2] == 0) {
1182     for (dim_t i=1; i<myN0-1; i++) {
1183     index.push_back(i);
1184     index.push_back(i+1);
1185     index.push_back(i+2);
1186     ptr.push_back(ptr.back()+3);
1187     }
1188     index.push_back(myN0-1);
1189     index.push_back(myN0);
1190     n+=2;
1191     if (faces[1] == 0) {
1192     index.push_back(myN0+1);
1193     index.push_back(myN0+2);
1194     index.push_back(myN0+3);
1195     n+=3;
1196     }
1197     } else {
1198     for (dim_t i=1; i<myN0-1; i++) {
1199     ptr.push_back(ptr.back());
1200     }
1201     if (faces[1] == 0) {
1202     index.push_back(myN0+2);
1203     index.push_back(myN0+3);
1204     n+=2;
1205     }
1206     }
1207     // n=neighbours of bottom-right corner node
1208     ptr.push_back(ptr.back()+n);
1209    
1210     // 2nd row to 2nd last row
1211     for (dim_t i=1; i<myN1-1; i++) {
1212     // left edge
1213     if (faces[0] == 0) {
1214     index.push_back(2*(myN0+myN1+2)-i);
1215     index.push_back(2*(myN0+myN1+2)-i-1);
1216     index.push_back(2*(myN0+myN1+2)-i-2);
1217     ptr.push_back(ptr.back()+3);
1218     } else {
1219     ptr.push_back(ptr.back());
1220     }
1221     for (dim_t j=1; j<myN0-1; j++) {
1222     ptr.push_back(ptr.back());
1223     }
1224     // right edge
1225     if (faces[1] == 0) {
1226     index.push_back(myN0+i+1);
1227     index.push_back(myN0+i+2);
1228     index.push_back(myN0+i+3);
1229     ptr.push_back(ptr.back()+3);
1230     } else {
1231     ptr.push_back(ptr.back());
1232     }
1233     }
1234    
1235     // top edge
1236     n=0;
1237     if (faces[0] == 0) {
1238     index.push_back(2*myN0+myN1+5);
1239     index.push_back(2*myN0+myN1+4);
1240     n+=2;
1241     if (faces[3] == 0) {
1242     index.push_back(2*myN0+myN1+3);
1243     index.push_back(2*myN0+myN1+2);
1244     index.push_back(2*myN0+myN1+1);
1245     n+=3;
1246     }
1247     } else if (faces[3] == 0) {
1248     index.push_back(2*myN0+myN1+2);
1249     index.push_back(2*myN0+myN1+1);
1250     n+=2;
1251     }
1252     // n=neighbours of top-left corner node
1253     ptr.push_back(ptr.back()+n);
1254     n=0;
1255     if (faces[3] == 0) {
1256     for (dim_t i=1; i<myN0-1; i++) {
1257     index.push_back(2*myN0+myN1+i+1);
1258     index.push_back(2*myN0+myN1+i);
1259     index.push_back(2*myN0+myN1+i-1);
1260     ptr.push_back(ptr.back()+3);
1261     }
1262     index.push_back(myN0+myN1+4);
1263     index.push_back(myN0+myN1+3);
1264     n+=2;
1265     if (faces[1] == 0) {
1266     index.push_back(myN0+myN1+2);
1267     index.push_back(myN0+myN1+1);
1268     index.push_back(myN0+myN1);
1269     n+=3;
1270     }
1271     } else {
1272     for (dim_t i=1; i<myN0-1; i++) {
1273     ptr.push_back(ptr.back());
1274     }
1275     if (faces[1] == 0) {
1276     index.push_back(myN0+myN1+1);
1277     index.push_back(myN0+myN1);
1278     n+=2;
1279     }
1280     }
1281     // n=neighbours of top-right corner node
1282     ptr.push_back(ptr.back()+n);
1283    
1284     dim_t M=ptr.size()-1;
1285     map<index_t,index_t> idMap;
1286     dim_t N=0;
1287     for (IndexVector::iterator it=index.begin(); it!=index.end(); it++) {
1288     if (idMap.find(*it)==idMap.end()) {
1289     idMap[*it]=N;
1290     *it=N++;
1291     } else {
1292     *it=idMap[*it];
1293     }
1294     }
1295    
1296 caltinay 3702 /*
1297 caltinay 3699 cout << "--- colCouple_pattern ---" << endl;
1298     cout << "M=" << M << ", N=" << N << endl;
1299     for (size_t i=0; i<ptr.size(); i++) {
1300     cout << "ptr[" << i << "]=" << ptr[i] << endl;
1301     }
1302     for (size_t i=0; i<index.size(); i++) {
1303     cout << "index[" << i << "]=" << index[i] << endl;
1304     }
1305 caltinay 3702 */
1306 caltinay 3699
1307     // now build the row couple pattern
1308     IndexVector ptr2(1,0);
1309     IndexVector index2;
1310     for (dim_t id=0; id<N; id++) {
1311     n=0;
1312     for (dim_t i=0; i<M; i++) {
1313     for (dim_t j=ptr[i]; j<ptr[i+1]; j++) {
1314     if (index[j]==id) {
1315     index2.push_back(i);
1316     n++;
1317     break;
1318     }
1319     }
1320     }
1321     ptr2.push_back(ptr2.back()+n);
1322     }
1323    
1324 caltinay 3702 /*
1325 caltinay 3699 cout << "--- rowCouple_pattern ---" << endl;
1326     cout << "M=" << N << ", N=" << M << endl;
1327     for (size_t i=0; i<ptr2.size(); i++) {
1328     cout << "ptr[" << i << "]=" << ptr2[i] << endl;
1329     }
1330     for (size_t i=0; i<index2.size(); i++) {
1331     cout << "index[" << i << "]=" << index2[i] << endl;
1332     }
1333 caltinay 3702 */
1334 caltinay 3699
1335     // paso will manage the memory
1336     index_t* indexC = MEMALLOC(index.size(), index_t);
1337     index_t* ptrC = MEMALLOC(ptr.size(), index_t);
1338     copy(index.begin(), index.end(), indexC);
1339     copy(ptr.begin(), ptr.end(), ptrC);
1340     *colPattern=Paso_Pattern_alloc(MATRIX_FORMAT_DEFAULT, M, N, ptrC, indexC);
1341    
1342     // paso will manage the memory
1343     indexC = MEMALLOC(index2.size(), index_t);
1344     ptrC = MEMALLOC(ptr2.size(), index_t);
1345     copy(index2.begin(), index2.end(), indexC);
1346     copy(ptr2.begin(), ptr2.end(), ptrC);
1347     *rowPattern=Paso_Pattern_alloc(MATRIX_FORMAT_DEFAULT, N, M, ptrC, indexC);
1348     }
1349    
1350 caltinay 3702 //protected
1351 caltinay 3711 void Rectangle::interpolateNodesOnElements(escript::Data& out,
1352     escript::Data& in, bool reduced) const
1353 caltinay 3702 {
1354     const dim_t numComp = in.getDataPointSize();
1355 caltinay 3711 if (reduced) {
1356 caltinay 3724 /*** GENERATOR SNIP_INTERPOLATE_REDUCED_ELEMENTS TOP */
1357     const double c0 = .25;
1358 caltinay 3702 #pragma omp parallel for
1359 caltinay 3711 for (index_t k1=0; k1 < m_NE1; ++k1) {
1360     for (index_t k0=0; k0 < m_NE0; ++k0) {
1361     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,k1+1, m_N0));
1362     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,k1, m_N0));
1363     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,k1, m_N0));
1364     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,k1+1, m_N0));
1365     double* o = out.getSampleDataRW(INDEX2(k0,k1,m_NE0));
1366     for (index_t i=0; i < numComp; ++i) {
1367 caltinay 3724 o[INDEX2(i,numComp,0)] = c0*(f_00[i] + f_01[i] + f_10[i] + f_11[i]);
1368 caltinay 3711 } /* end of component loop i */
1369     } /* end of k0 loop */
1370     } /* end of k1 loop */
1371     /* GENERATOR SNIP_INTERPOLATE_REDUCED_ELEMENTS BOTTOM */
1372     } else {
1373 caltinay 3724 /*** GENERATOR SNIP_INTERPOLATE_ELEMENTS TOP */
1374     const double c0 = .16666666666666666667;
1375     const double c1 = .044658198738520451079;
1376     const double c2 = .62200846792814621559;
1377 caltinay 3711 #pragma omp parallel for
1378     for (index_t k1=0; k1 < m_NE1; ++k1) {
1379     for (index_t k0=0; k0 < m_NE0; ++k0) {
1380     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,k1, m_N0));
1381     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,k1+1, m_N0));
1382     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,k1+1, m_N0));
1383     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,k1, m_N0));
1384     double* o = out.getSampleDataRW(INDEX2(k0,k1,m_NE0));
1385     for (index_t i=0; i < numComp; ++i) {
1386 caltinay 3724 o[INDEX2(i,numComp,0)] = f_00[i]*c2 + f_11[i]*c1 + c0*(f_01[i] + f_10[i]);
1387     o[INDEX2(i,numComp,1)] = f_01[i]*c1 + f_10[i]*c2 + c0*(f_00[i] + f_11[i]);
1388     o[INDEX2(i,numComp,2)] = f_01[i]*c2 + f_10[i]*c1 + c0*(f_00[i] + f_11[i]);
1389     o[INDEX2(i,numComp,3)] = f_00[i]*c1 + f_11[i]*c2 + c0*(f_01[i] + f_10[i]);
1390 caltinay 3711 } /* end of component loop i */
1391     } /* end of k0 loop */
1392     } /* end of k1 loop */
1393     /* GENERATOR SNIP_INTERPOLATE_ELEMENTS BOTTOM */
1394     }
1395 caltinay 3702 }
1396    
1397     //protected
1398 caltinay 3711 void Rectangle::interpolateNodesOnFaces(escript::Data& out, escript::Data& in,
1399     bool reduced) const
1400 caltinay 3702 {
1401 caltinay 3704 const dim_t numComp = in.getDataPointSize();
1402 caltinay 3711 if (reduced) {
1403 caltinay 3724 const double c0 = .5;
1404     #pragma omp parallel
1405     {
1406     /*** GENERATOR SNIP_INTERPOLATE_REDUCED_FACES TOP */
1407     if (m_faceOffset[0] > -1) {
1408     #pragma omp for nowait
1409     for (index_t k1=0; k1 < m_NE1; ++k1) {
1410     const register double* f_00 = in.getSampleDataRO(INDEX2(0,k1, m_N0));
1411     const register double* f_01 = in.getSampleDataRO(INDEX2(0,k1+1, m_N0));
1412     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
1413     for (index_t i=0; i < numComp; ++i) {
1414     o[INDEX2(i,numComp,0)] = c0*(f_00[i] + f_01[i]);
1415     } /* end of component loop i */
1416     } /* end of k1 loop */
1417     } /* end of face 0 */
1418     if (m_faceOffset[1] > -1) {
1419     #pragma omp for nowait
1420     for (index_t k1=0; k1 < m_NE1; ++k1) {
1421     const register double* f_10 = in.getSampleDataRO(INDEX2(m_N0-1,k1, m_N0));
1422     const register double* f_11 = in.getSampleDataRO(INDEX2(m_N0-1,k1+1, m_N0));
1423     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
1424     for (index_t i=0; i < numComp; ++i) {
1425     o[INDEX2(i,numComp,0)] = c0*(f_10[i] + f_11[i]);
1426     } /* end of component loop i */
1427     } /* end of k1 loop */
1428     } /* end of face 1 */
1429     if (m_faceOffset[2] > -1) {
1430     #pragma omp for nowait
1431     for (index_t k0=0; k0 < m_NE0; ++k0) {
1432     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,0, m_N0));
1433     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,0, m_N0));
1434     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
1435     for (index_t i=0; i < numComp; ++i) {
1436     o[INDEX2(i,numComp,0)] = c0*(f_00[i] + f_10[i]);
1437     } /* end of component loop i */
1438     } /* end of k0 loop */
1439     } /* end of face 2 */
1440     if (m_faceOffset[3] > -1) {
1441     #pragma omp for nowait
1442     for (index_t k0=0; k0 < m_NE0; ++k0) {
1443     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,m_N1-1, m_N0));
1444     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,m_N1-1, m_N0));
1445     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
1446     for (index_t i=0; i < numComp; ++i) {
1447     o[INDEX2(i,numComp,0)] = c0*(f_01[i] + f_11[i]);
1448     } /* end of component loop i */
1449     } /* end of k0 loop */
1450     } /* end of face 3 */
1451     /* GENERATOR SNIP_INTERPOLATE_REDUCED_FACES BOTTOM */
1452     } // end of parallel section
1453 caltinay 3711 } else {
1454 caltinay 3724 const double c0 = 0.21132486540518711775;
1455     const double c1 = 0.78867513459481288225;
1456     #pragma omp parallel
1457     {
1458     /*** GENERATOR SNIP_INTERPOLATE_FACES TOP */
1459     if (m_faceOffset[0] > -1) {
1460     #pragma omp for nowait
1461     for (index_t k1=0; k1 < m_NE1; ++k1) {
1462     const register double* f_01 = in.getSampleDataRO(INDEX2(0,k1+1, m_N0));
1463     const register double* f_00 = in.getSampleDataRO(INDEX2(0,k1, m_N0));
1464     double* o = out.getSampleDataRW(m_faceOffset[0]+k1);
1465     for (index_t i=0; i < numComp; ++i) {
1466     o[INDEX2(i,numComp,0)] = f_00[i]*c1 + f_01[i]*c0;
1467     o[INDEX2(i,numComp,1)] = f_00[i]*c0 + f_01[i]*c1;
1468     } /* end of component loop i */
1469     } /* end of k1 loop */
1470     } /* end of face 0 */
1471     if (m_faceOffset[1] > -1) {
1472     #pragma omp for nowait
1473     for (index_t k1=0; k1 < m_NE1; ++k1) {
1474     const register double* f_10 = in.getSampleDataRO(INDEX2(m_N0-1,k1, m_N0));
1475     const register double* f_11 = in.getSampleDataRO(INDEX2(m_N0-1,k1+1, m_N0));
1476     double* o = out.getSampleDataRW(m_faceOffset[1]+k1);
1477     for (index_t i=0; i < numComp; ++i) {
1478     o[INDEX2(i,numComp,0)] = f_10[i]*c1 + f_11[i]*c0;
1479     o[INDEX2(i,numComp,1)] = f_10[i]*c0 + f_11[i]*c1;
1480     } /* end of component loop i */
1481     } /* end of k1 loop */
1482     } /* end of face 1 */
1483     if (m_faceOffset[2] > -1) {
1484     #pragma omp for nowait
1485     for (index_t k0=0; k0 < m_NE0; ++k0) {
1486     const register double* f_10 = in.getSampleDataRO(INDEX2(k0+1,0, m_N0));
1487     const register double* f_00 = in.getSampleDataRO(INDEX2(k0,0, m_N0));
1488     double* o = out.getSampleDataRW(m_faceOffset[2]+k0);
1489     for (index_t i=0; i < numComp; ++i) {
1490     o[INDEX2(i,numComp,0)] = f_00[i]*c1 + f_10[i]*c0;
1491     o[INDEX2(i,numComp,1)] = f_00[i]*c0 + f_10[i]*c1;
1492     } /* end of component loop i */
1493     } /* end of k0 loop */
1494     } /* end of face 2 */
1495     if (m_faceOffset[3] > -1) {
1496     #pragma omp for nowait
1497     for (index_t k0=0; k0 < m_NE0; ++k0) {
1498     const register double* f_11 = in.getSampleDataRO(INDEX2(k0+1,m_N1-1, m_N0));
1499     const register double* f_01 = in.getSampleDataRO(INDEX2(k0,m_N1-1, m_N0));
1500     double* o = out.getSampleDataRW(m_faceOffset[3]+k0);
1501     for (index_t i=0; i < numComp; ++i) {
1502     o[INDEX2(i,numComp,0)] = f_01[i]*c1 + f_11[i]*c0;
1503     o[INDEX2(i,numComp,1)] = f_01[i]*c0 + f_11[i]*c1;
1504     } /* end of component loop i */
1505     } /* end of k0 loop */
1506     } /* end of face 3 */
1507     /* GENERATOR SNIP_INTERPOLATE_FACES BOTTOM */
1508     } // end of parallel section
1509 caltinay 3711 }
1510 caltinay 3702 }
1511    
1512 caltinay 3748 //protected
1513     void Rectangle::assemblePDESingle(Paso_SystemMatrix* mat,
1514     escript::Data& rhs, const escript::Data& A, const escript::Data& B,
1515     const escript::Data& C, const escript::Data& D,
1516     const escript::Data& X, const escript::Data& Y,
1517     const escript::Data& d, const escript::Data& y) const
1518     {
1519     const double h0 = m_l0/m_gNE0;
1520     const double h1 = m_l1/m_gNE1;
1521     /* GENERATOR SNIP_PDE_SINGLE_PRE TOP */
1522     const double w0 = -0.1555021169820365539*h1/h0;
1523     const double w1 = 0.041666666666666666667;
1524     const double w10 = -0.041666666666666666667*h0/h1;
1525     const double w11 = 0.1555021169820365539*h1/h0;
1526     const double w12 = 0.1555021169820365539*h0/h1;
1527     const double w13 = 0.01116454968463011277*h0/h1;
1528     const double w14 = 0.01116454968463011277*h1/h0;
1529     const double w15 = 0.041666666666666666667*h1/h0;
1530     const double w16 = -0.01116454968463011277*h0/h1;
1531     const double w17 = -0.1555021169820365539*h0/h1;
1532     const double w18 = -0.33333333333333333333*h1/h0;
1533     const double w19 = 0.25000000000000000000;
1534     const double w2 = -0.15550211698203655390;
1535     const double w20 = -0.25000000000000000000;
1536     const double w21 = 0.16666666666666666667*h0/h1;
1537     const double w22 = -0.16666666666666666667*h1/h0;
1538     const double w23 = -0.16666666666666666667*h0/h1;
1539     const double w24 = 0.33333333333333333333*h1/h0;
1540     const double w25 = 0.33333333333333333333*h0/h1;
1541     const double w26 = 0.16666666666666666667*h1/h0;
1542     const double w27 = -0.33333333333333333333*h0/h1;
1543     const double w28 = -0.032861463941450536761*h1;
1544     const double w29 = -0.032861463941450536761*h0;
1545     const double w3 = 0.041666666666666666667*h0/h1;
1546     const double w30 = -0.12264065304058601714*h1;
1547     const double w31 = -0.0023593469594139828636*h1;
1548     const double w32 = -0.008805202725216129906*h0;
1549     const double w33 = -0.008805202725216129906*h1;
1550     const double w34 = 0.032861463941450536761*h1;
1551     const double w35 = 0.008805202725216129906*h1;
1552     const double w36 = 0.008805202725216129906*h0;
1553     const double w37 = 0.0023593469594139828636*h1;
1554     const double w38 = 0.12264065304058601714*h1;
1555     const double w39 = 0.032861463941450536761*h0;
1556     const double w4 = 0.15550211698203655390;
1557     const double w40 = -0.12264065304058601714*h0;
1558     const double w41 = -0.0023593469594139828636*h0;
1559     const double w42 = 0.0023593469594139828636*h0;
1560     const double w43 = 0.12264065304058601714*h0;
1561     const double w44 = -0.16666666666666666667*h1;
1562     const double w45 = -0.083333333333333333333*h0;
1563     const double w46 = 0.083333333333333333333*h1;
1564     const double w47 = 0.16666666666666666667*h1;
1565     const double w48 = 0.083333333333333333333*h0;
1566     const double w49 = -0.16666666666666666667*h0;
1567     const double w5 = -0.041666666666666666667;
1568     const double w50 = 0.16666666666666666667*h0;
1569     const double w51 = -0.083333333333333333333*h1;
1570     const double w52 = 0.025917019497006092316*h0*h1;
1571     const double w53 = 0.0018607582807716854616*h0*h1;
1572     const double w54 = 0.0069444444444444444444*h0*h1;
1573     const double w55 = 0.09672363354357992482*h0*h1;
1574     const double w56 = 0.00049858867864229740201*h0*h1;
1575     const double w57 = 0.055555555555555555556*h0*h1;
1576     const double w58 = 0.027777777777777777778*h0*h1;
1577     const double w59 = 0.11111111111111111111*h0*h1;
1578     const double w6 = -0.01116454968463011277*h1/h0;
1579     const double w60 = -0.19716878364870322056*h1;
1580     const double w61 = -0.19716878364870322056*h0;
1581     const double w62 = -0.052831216351296779436*h0;
1582     const double w63 = -0.052831216351296779436*h1;
1583     const double w64 = 0.19716878364870322056*h1;
1584     const double w65 = 0.052831216351296779436*h1;
1585     const double w66 = 0.19716878364870322056*h0;
1586     const double w67 = 0.052831216351296779436*h0;
1587     const double w68 = -0.5*h1;
1588     const double w69 = -0.5*h0;
1589     const double w7 = 0.011164549684630112770;
1590     const double w70 = 0.5*h1;
1591     const double w71 = 0.5*h0;
1592     const double w72 = 0.1555021169820365539*h0*h1;
1593     const double w73 = 0.041666666666666666667*h0*h1;
1594     const double w74 = 0.01116454968463011277*h0*h1;
1595     const double w75 = 0.25*h0*h1;
1596     const double w8 = -0.011164549684630112770;
1597     const double w9 = -0.041666666666666666667*h1/h0;
1598     /* GENERATOR SNIP_PDE_SINGLE_PRE BOTTOM */
1599    
1600     rhs.requireWrite();
1601     #pragma omp parallel
1602     {
1603     for (index_t k1_0=0; k1_0<2; k1_0++) { // coloring
1604     #pragma omp for
1605     for (index_t k1=k1_0; k1<m_NE1; k1+=2) {
1606     for (index_t k0=0; k0<m_NE0; ++k0) {
1607     bool add_EM_S=false;
1608     bool add_EM_F=false;
1609     vector<double> EM_S(4*4, 0);
1610     vector<double> EM_F(4, 0);
1611     const index_t e = k0 + m_NE0*k1;
1612     /* GENERATOR SNIP_PDE_SINGLE TOP */
1613     ///////////////
1614     // process A //
1615     ///////////////
1616     if (!A.isEmpty()) {
1617     add_EM_S=true;
1618     const double* A_p=const_cast<escript::Data*>(&A)->getSampleDataRO(e);
1619     if (A.actsExpanded()) {
1620     const register double A_00_0 = A_p[INDEX3(0,0,0,2,2)];
1621     const register double A_01_0 = A_p[INDEX3(0,1,0,2,2)];
1622     const register double A_10_0 = A_p[INDEX3(1,0,0,2,2)];
1623     const register double A_11_0 = A_p[INDEX3(1,1,0,2,2)];
1624     const register double A_00_1 = A_p[INDEX3(0,0,1,2,2)];
1625     const register double A_01_1 = A_p[INDEX3(0,1,1,2,2)];
1626     const register double A_10_1 = A_p[INDEX3(1,0,1,2,2)];
1627     const register double A_11_1 = A_p[INDEX3(1,1,1,2,2)];
1628     const register double A_00_2 = A_p[INDEX3(0,0,2,2,2)];
1629     const register double A_01_2 = A_p[INDEX3(0,1,2,2,2)];
1630     const register double A_10_2 = A_p[INDEX3(1,0,2,2,2)];
1631     const register double A_11_2 = A_p[INDEX3(1,1,2,2,2)];
1632     const register double A_00_3 = A_p[INDEX3(0,0,3,2,2)];
1633     const register double A_01_3 = A_p[INDEX3(0,1,3,2,2)];
1634     const register double A_10_3 = A_p[INDEX3(1,0,3,2,2)];
1635     const register double A_11_3 = A_p[INDEX3(1,1,3,2,2)];
1636     const register double tmp4_0 = A_10_1 + A_10_2;
1637     const register double tmp12_0 = A_11_0 + A_11_2;
1638     const register double tmp2_0 = A_11_0 + A_11_1 + A_11_2 + A_11_3;
1639     const register double tmp10_0 = A_01_3 + A_10_3;
1640     const register double tmp14_0 = A_01_0 + A_01_3 + A_10_0 + A_10_3;
1641     const register double tmp0_0 = A_01_0 + A_01_3;
1642     const register double tmp13_0 = A_01_2 + A_10_1;
1643     const register double tmp3_0 = A_00_2 + A_00_3;
1644     const register double tmp11_0 = A_11_1 + A_11_3;
1645     const register double tmp18_0 = A_01_1 + A_10_1;
1646     const register double tmp1_0 = A_00_0 + A_00_1;
1647     const register double tmp15_0 = A_01_1 + A_10_2;
1648     const register double tmp5_0 = A_00_0 + A_00_1 + A_00_2 + A_00_3;
1649     const register double tmp16_0 = A_10_0 + A_10_3;
1650     const register double tmp6_0 = A_01_3 + A_10_0;
1651     const register double tmp17_0 = A_01_1 + A_01_2;
1652     const register double tmp9_0 = A_01_0 + A_10_0;
1653     const register double tmp7_0 = A_01_0 + A_10_3;
1654     const register double tmp8_0 = A_01_1 + A_01_2 + A_10_1 + A_10_2;
1655     const register double tmp19_0 = A_01_2 + A_10_2;
1656     const register double tmp14_1 = A_10_0*w8;
1657     const register double tmp23_1 = tmp3_0*w14;
1658     const register double tmp35_1 = A_01_0*w8;
1659     const register double tmp54_1 = tmp13_0*w8;
1660     const register double tmp20_1 = tmp9_0*w4;
1661     const register double tmp25_1 = tmp12_0*w12;
1662     const register double tmp2_1 = A_01_1*w4;
1663     const register double tmp44_1 = tmp7_0*w7;
1664     const register double tmp26_1 = tmp10_0*w4;
1665     const register double tmp52_1 = tmp18_0*w8;
1666     const register double tmp48_1 = A_10_1*w7;
1667     const register double tmp46_1 = A_01_3*w8;
1668     const register double tmp50_1 = A_01_0*w2;
1669     const register double tmp8_1 = tmp4_0*w5;
1670     const register double tmp56_1 = tmp19_0*w8;
1671     const register double tmp9_1 = tmp2_0*w10;
1672     const register double tmp19_1 = A_10_3*w2;
1673     const register double tmp47_1 = A_10_2*w4;
1674     const register double tmp16_1 = tmp3_0*w0;
1675     const register double tmp18_1 = tmp1_0*w6;
1676     const register double tmp31_1 = tmp11_0*w12;
1677     const register double tmp55_1 = tmp15_0*w2;
1678     const register double tmp39_1 = A_10_2*w7;
1679     const register double tmp11_1 = tmp6_0*w7;
1680     const register double tmp40_1 = tmp11_0*w17;
1681     const register double tmp34_1 = tmp15_0*w8;
1682     const register double tmp33_1 = tmp14_0*w5;
1683     const register double tmp24_1 = tmp11_0*w13;
1684     const register double tmp3_1 = tmp1_0*w0;
1685     const register double tmp5_1 = tmp2_0*w3;
1686     const register double tmp43_1 = tmp17_0*w5;
1687     const register double tmp15_1 = A_01_2*w4;
1688     const register double tmp53_1 = tmp19_0*w2;
1689     const register double tmp27_1 = tmp3_0*w11;
1690     const register double tmp32_1 = tmp13_0*w2;
1691     const register double tmp10_1 = tmp5_0*w9;
1692     const register double tmp37_1 = A_10_1*w4;
1693     const register double tmp38_1 = tmp5_0*w15;
1694     const register double tmp17_1 = A_01_1*w7;
1695     const register double tmp12_1 = tmp7_0*w4;
1696     const register double tmp22_1 = tmp10_0*w7;
1697     const register double tmp57_1 = tmp18_0*w2;
1698     const register double tmp28_1 = tmp9_0*w7;
1699     const register double tmp29_1 = tmp1_0*w14;
1700     const register double tmp51_1 = tmp11_0*w16;
1701     const register double tmp42_1 = tmp12_0*w16;
1702     const register double tmp49_1 = tmp12_0*w17;
1703     const register double tmp21_1 = tmp1_0*w11;
1704     const register double tmp1_1 = tmp0_0*w1;
1705     const register double tmp45_1 = tmp6_0*w4;
1706     const register double tmp7_1 = A_10_0*w2;
1707     const register double tmp6_1 = tmp3_0*w6;
1708     const register double tmp13_1 = tmp8_0*w1;
1709     const register double tmp36_1 = tmp16_0*w1;
1710     const register double tmp41_1 = A_01_3*w2;
1711     const register double tmp30_1 = tmp12_0*w13;
1712     const register double tmp4_1 = A_01_2*w7;
1713     const register double tmp0_1 = A_10_3*w8;
1714     EM_S[INDEX2(0,1,4)]+=tmp0_1 + tmp1_1 + tmp2_1 + tmp3_1 + tmp4_1 + tmp5_1 + tmp6_1 + tmp7_1 + tmp8_1;
1715     EM_S[INDEX2(1,2,4)]+=tmp10_1 + tmp11_1 + tmp12_1 + tmp13_1 + tmp9_1;
1716     EM_S[INDEX2(3,2,4)]+=tmp14_1 + tmp15_1 + tmp16_1 + tmp17_1 + tmp18_1 + tmp19_1 + tmp1_1 + tmp5_1 + tmp8_1;
1717     EM_S[INDEX2(0,0,4)]+=tmp13_1 + tmp20_1 + tmp21_1 + tmp22_1 + tmp23_1 + tmp24_1 + tmp25_1;
1718     EM_S[INDEX2(3,3,4)]+=tmp13_1 + tmp26_1 + tmp27_1 + tmp28_1 + tmp29_1 + tmp30_1 + tmp31_1;
1719     EM_S[INDEX2(3,0,4)]+=tmp10_1 + tmp32_1 + tmp33_1 + tmp34_1 + tmp9_1;
1720     EM_S[INDEX2(3,1,4)]+=tmp35_1 + tmp36_1 + tmp37_1 + tmp38_1 + tmp39_1 + tmp40_1 + tmp41_1 + tmp42_1 + tmp43_1;
1721     EM_S[INDEX2(2,1,4)]+=tmp10_1 + tmp13_1 + tmp44_1 + tmp45_1 + tmp9_1;
1722     EM_S[INDEX2(0,2,4)]+=tmp36_1 + tmp38_1 + tmp43_1 + tmp46_1 + tmp47_1 + tmp48_1 + tmp49_1 + tmp50_1 + tmp51_1;
1723     EM_S[INDEX2(2,0,4)]+=tmp0_1 + tmp15_1 + tmp17_1 + tmp1_1 + tmp38_1 + tmp49_1 + tmp51_1 + tmp7_1 + tmp8_1;
1724     EM_S[INDEX2(1,3,4)]+=tmp14_1 + tmp19_1 + tmp1_1 + tmp2_1 + tmp38_1 + tmp40_1 + tmp42_1 + tmp4_1 + tmp8_1;
1725     EM_S[INDEX2(2,3,4)]+=tmp16_1 + tmp18_1 + tmp35_1 + tmp36_1 + tmp41_1 + tmp43_1 + tmp47_1 + tmp48_1 + tmp5_1;
1726     EM_S[INDEX2(2,2,4)]+=tmp24_1 + tmp25_1 + tmp27_1 + tmp29_1 + tmp33_1 + tmp52_1 + tmp53_1;
1727     EM_S[INDEX2(1,0,4)]+=tmp36_1 + tmp37_1 + tmp39_1 + tmp3_1 + tmp43_1 + tmp46_1 + tmp50_1 + tmp5_1 + tmp6_1;
1728     EM_S[INDEX2(0,3,4)]+=tmp10_1 + tmp33_1 + tmp54_1 + tmp55_1 + tmp9_1;
1729     EM_S[INDEX2(1,1,4)]+=tmp21_1 + tmp23_1 + tmp30_1 + tmp31_1 + tmp33_1 + tmp56_1 + tmp57_1;
1730     } else { /* constant data */
1731     const register double A_00 = A_p[INDEX2(0,0,2)];
1732     const register double A_01 = A_p[INDEX2(0,1,2)];
1733     const register double A_10 = A_p[INDEX2(1,0,2)];
1734     const register double A_11 = A_p[INDEX2(1,1,2)];
1735     const register double tmp0_0 = A_01 + A_10;
1736     const register double tmp0_1 = A_00*w18;
1737     const register double tmp10_1 = A_01*w20;
1738     const register double tmp12_1 = A_00*w26;
1739     const register double tmp4_1 = A_00*w22;
1740     const register double tmp8_1 = A_00*w24;
1741     const register double tmp13_1 = A_10*w19;
1742     const register double tmp9_1 = tmp0_0*w20;
1743     const register double tmp3_1 = A_11*w21;
1744     const register double tmp11_1 = A_11*w27;
1745     const register double tmp1_1 = A_01*w19;
1746     const register double tmp6_1 = A_11*w23;
1747     const register double tmp7_1 = A_11*w25;
1748     const register double tmp2_1 = A_10*w20;
1749     const register double tmp5_1 = tmp0_0*w19;
1750     EM_S[INDEX2(0,1,4)]+=tmp0_1 + tmp1_1 + tmp2_1 + tmp3_1;
1751     EM_S[INDEX2(1,2,4)]+=tmp4_1 + tmp5_1 + tmp6_1;
1752     EM_S[INDEX2(3,2,4)]+=tmp0_1 + tmp1_1 + tmp2_1 + tmp3_1;
1753     EM_S[INDEX2(0,0,4)]+=tmp5_1 + tmp7_1 + tmp8_1;
1754     EM_S[INDEX2(3,3,4)]+=tmp5_1 + tmp7_1 + tmp8_1;
1755     EM_S[INDEX2(3,0,4)]+=tmp4_1 + tmp6_1 + tmp9_1;
1756     EM_S[INDEX2(3,1,4)]+=tmp10_1 + tmp11_1 + tmp12_1 + tmp13_1;
1757     EM_S[INDEX2(2,1,4)]+=tmp4_1 + tmp5_1 + tmp6_1;
1758     EM_S[INDEX2(0,2,4)]+=tmp10_1 + tmp11_1 + tmp12_1 + tmp13_1;
1759     EM_S[INDEX2(2,0,4)]+=tmp11_1 + tmp12_1 + tmp1_1 + tmp2_1;
1760     EM_S[INDEX2(1,3,4)]+=tmp11_1 + tmp12_1 + tmp1_1 + tmp2_1;
1761     EM_S[INDEX2(2,3,4)]+=tmp0_1 + tmp10_1 + tmp13_1 + tmp3_1;
1762     EM_S[INDEX2(2,2,4)]+=tmp7_1 + tmp8_1 + tmp9_1;
1763     EM_S[INDEX2(1,0,4)]+=tmp0_1 + tmp10_1 + tmp13_1 + tmp3_1;
1764     EM_S[INDEX2(0,3,4)]+=tmp4_1 + tmp6_1 + tmp9_1;
1765     EM_S[INDEX2(1,1,4)]+=tmp7_1 + tmp8_1 + tmp9_1;
1766     }
1767     }
1768     ///////////////
1769     // process B //
1770     ///////////////
1771     if (!B.isEmpty()) {
1772     add_EM_S=true;
1773     const double* B_p=const_cast<escript::Data*>(&B)->getSampleDataRO(e);
1774     if (B.actsExpanded()) {
1775     const register double B_0_0 = B_p[INDEX2(0,0,2)];
1776     const register double B_1_0 = B_p[INDEX2(1,0,2)];
1777     const register double B_0_1 = B_p[INDEX2(0,1,2)];
1778     const register double B_1_1 = B_p[INDEX2(1,1,2)];
1779     const register double B_0_2 = B_p[INDEX2(0,2,2)];
1780     const register double B_1_2 = B_p[INDEX2(1,2,2)];
1781     const register double B_0_3 = B_p[INDEX2(0,3,2)];
1782     const register double B_1_3 = B_p[INDEX2(1,3,2)];
1783     const register double tmp3_0 = B_0_0 + B_0_2;
1784     const register double tmp1_0 = B_1_2 + B_1_3;
1785     const register double tmp2_0 = B_0_1 + B_0_3;
1786     const register double tmp0_0 = B_1_0 + B_1_1;
1787     const register double tmp63_1 = B_1_1*w42;
1788     const register double tmp79_1 = B_1_1*w40;
1789     const register double tmp37_1 = tmp3_0*w35;
1790     const register double tmp8_1 = tmp0_0*w32;
1791     const register double tmp71_1 = B_0_1*w34;
1792     const register double tmp19_1 = B_0_3*w31;
1793     const register double tmp15_1 = B_0_3*w34;
1794     const register double tmp9_1 = tmp3_0*w34;
1795     const register double tmp35_1 = B_1_0*w36;
1796     const register double tmp66_1 = B_0_3*w28;
1797     const register double tmp28_1 = B_1_0*w42;
1798     const register double tmp22_1 = B_1_0*w40;
1799     const register double tmp16_1 = B_1_2*w29;
1800     const register double tmp6_1 = tmp2_0*w35;
1801     const register double tmp55_1 = B_1_3*w40;
1802     const register double tmp50_1 = B_1_3*w42;
1803     const register double tmp7_1 = tmp1_0*w29;
1804     const register double tmp1_1 = tmp1_0*w32;
1805     const register double tmp57_1 = B_0_3*w30;
1806     const register double tmp18_1 = B_1_1*w32;
1807     const register double tmp53_1 = B_1_0*w41;
1808     const register double tmp61_1 = B_1_3*w36;
1809     const register double tmp27_1 = B_0_3*w38;
1810     const register double tmp64_1 = B_0_2*w30;
1811     const register double tmp76_1 = B_0_1*w38;
1812     const register double tmp39_1 = tmp2_0*w34;
1813     const register double tmp62_1 = B_0_1*w31;
1814     const register double tmp56_1 = B_0_0*w31;
1815     const register double tmp49_1 = B_1_1*w36;
1816     const register double tmp2_1 = B_0_2*w31;
1817     const register double tmp23_1 = B_0_2*w33;
1818     const register double tmp38_1 = B_1_1*w43;
1819     const register double tmp74_1 = B_1_2*w41;
1820     const register double tmp43_1 = B_1_1*w41;
1821     const register double tmp58_1 = B_0_2*w28;
1822     const register double tmp67_1 = B_0_0*w33;
1823     const register double tmp33_1 = tmp0_0*w39;
1824     const register double tmp4_1 = B_0_0*w28;
1825     const register double tmp20_1 = B_0_0*w30;
1826     const register double tmp13_1 = B_0_2*w38;
1827     const register double tmp65_1 = B_1_2*w43;
1828     const register double tmp0_1 = tmp0_0*w29;
1829     const register double tmp41_1 = tmp3_0*w33;
1830     const register double tmp73_1 = B_0_2*w37;
1831     const register double tmp69_1 = B_0_0*w38;
1832     const register double tmp48_1 = B_1_2*w39;
1833     const register double tmp59_1 = B_0_1*w33;
1834     const register double tmp17_1 = B_1_3*w41;
1835     const register double tmp5_1 = B_0_3*w33;
1836     const register double tmp3_1 = B_0_1*w30;
1837     const register double tmp21_1 = B_0_1*w28;
1838     const register double tmp42_1 = B_1_0*w29;
1839     const register double tmp54_1 = B_1_2*w32;
1840     const register double tmp60_1 = B_1_0*w39;
1841     const register double tmp32_1 = tmp1_0*w36;
1842     const register double tmp10_1 = B_0_1*w37;
1843     const register double tmp14_1 = B_0_0*w35;
1844     const register double tmp29_1 = B_0_1*w35;
1845     const register double tmp26_1 = B_1_2*w36;
1846     const register double tmp30_1 = B_1_3*w43;
1847     const register double tmp70_1 = B_0_2*w35;
1848     const register double tmp34_1 = B_1_3*w39;
1849     const register double tmp51_1 = B_1_0*w43;
1850     const register double tmp31_1 = B_0_2*w34;
1851     const register double tmp45_1 = tmp3_0*w28;
1852     const register double tmp11_1 = tmp1_0*w39;
1853     const register double tmp52_1 = B_1_1*w29;
1854     const register double tmp44_1 = B_1_3*w32;
1855     const register double tmp25_1 = B_1_1*w39;
1856     const register double tmp47_1 = tmp2_0*w33;
1857     const register double tmp72_1 = B_1_3*w29;
1858     const register double tmp40_1 = tmp2_0*w28;
1859     const register double tmp46_1 = B_1_2*w40;
1860     const register double tmp36_1 = B_1_2*w42;
1861     const register double tmp24_1 = B_0_0*w37;
1862     const register double tmp77_1 = B_0_3*w35;
1863     const register double tmp68_1 = B_0_3*w37;
1864     const register double tmp78_1 = B_0_0*w34;
1865     const register double tmp12_1 = tmp0_0*w36;
1866     const register double tmp75_1 = B_1_0*w32;
1867     EM_S[INDEX2(0,1,4)]+=tmp0_1 + tmp1_1 + tmp2_1 + tmp3_1 + tmp4_1 + tmp5_1;
1868     EM_S[INDEX2(1,2,4)]+=tmp6_1 + tmp7_1 + tmp8_1 + tmp9_1;
1869     EM_S[INDEX2(3,2,4)]+=tmp10_1 + tmp11_1 + tmp12_1 + tmp13_1 + tmp14_1 + tmp15_1;
1870     EM_S[INDEX2(0,0,4)]+=tmp16_1 + tmp17_1 + tmp18_1 + tmp19_1 + tmp20_1 + tmp21_1 + tmp22_1 + tmp23_1;
1871