1 | #include "mpi.hpp" |
---|
2 | #include <map> |
---|
3 | #include "cputime.hpp" /* time */ |
---|
4 | #include "meshutil.hpp" |
---|
5 | #include "polyg.hpp" |
---|
6 | #include "circle.hpp" |
---|
7 | #include "intersect.hpp" |
---|
8 | #include "intersection_ym.hpp" |
---|
9 | #include "errhandle.hpp" |
---|
10 | #include "mpi_routing.hpp" |
---|
11 | #include "gridRemap.hpp" |
---|
12 | |
---|
13 | #include "mapper.hpp" |
---|
14 | |
---|
15 | namespace sphereRemap { |
---|
16 | |
---|
17 | /* A subdivition of an array into N sub-arays |
---|
18 | can be represented by the length of the N arrays |
---|
19 | or by the offsets when each of the N arrays starts. |
---|
20 | This function convertes from the former to the latter. */ |
---|
21 | void cptOffsetsFromLengths(const int *lengths, int *offsets, int sz) |
---|
22 | { |
---|
23 | offsets[0] = 0; |
---|
24 | for (int i = 1; i < sz; i++) |
---|
25 | offsets[i] = offsets[i-1] + lengths[i-1]; |
---|
26 | } |
---|
27 | |
---|
28 | |
---|
29 | void Mapper::setSourceMesh(const double* boundsLon, const double* boundsLat, const double* area, int nVertex, int nbCells, const double* pole, const long int* globalId) |
---|
30 | { |
---|
31 | srcGrid.pole = Coord(pole[0], pole[1], pole[2]); |
---|
32 | |
---|
33 | int mpiRank, mpiSize; |
---|
34 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
35 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
36 | |
---|
37 | sourceElements.reserve(nbCells); |
---|
38 | sourceMesh.reserve(nbCells); |
---|
39 | sourceGlobalId.resize(nbCells) ; |
---|
40 | |
---|
41 | if (globalId==NULL) |
---|
42 | { |
---|
43 | long int offset ; |
---|
44 | long int nb=nbCells ; |
---|
45 | ep_lib::MPI_Scan(&nb,&offset,1,EP_LONG,EP_SUM,communicator) ; |
---|
46 | offset=offset-nb ; |
---|
47 | for(int i=0;i<nbCells;i++) sourceGlobalId[i]=offset+i ; |
---|
48 | } |
---|
49 | else sourceGlobalId.assign(globalId,globalId+nbCells); |
---|
50 | |
---|
51 | for (int i = 0; i < nbCells; i++) |
---|
52 | { |
---|
53 | int offs = i*nVertex; |
---|
54 | Elt elt(boundsLon + offs, boundsLat + offs, nVertex); |
---|
55 | elt.src_id.rank = mpiRank; |
---|
56 | elt.src_id.ind = i; |
---|
57 | elt.src_id.globalId = sourceGlobalId[i]; |
---|
58 | sourceElements.push_back(elt); |
---|
59 | sourceMesh.push_back(Node(elt.x, cptRadius(elt), &sourceElements.back())); |
---|
60 | cptEltGeom(sourceElements[i], Coord(pole[0], pole[1], pole[2])); |
---|
61 | if (area!=NULL) sourceElements[i].given_area=area[i] ; |
---|
62 | else sourceElements[i].given_area=sourceElements[i].area ; |
---|
63 | } |
---|
64 | |
---|
65 | } |
---|
66 | |
---|
67 | void Mapper::setTargetMesh(const double* boundsLon, const double* boundsLat, const double* area, int nVertex, int nbCells, const double* pole, const long int* globalId) |
---|
68 | { |
---|
69 | tgtGrid.pole = Coord(pole[0], pole[1], pole[2]); |
---|
70 | |
---|
71 | int mpiRank, mpiSize; |
---|
72 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
73 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
74 | |
---|
75 | targetElements.reserve(nbCells); |
---|
76 | targetMesh.reserve(nbCells); |
---|
77 | |
---|
78 | targetGlobalId.resize(nbCells) ; |
---|
79 | if (globalId==NULL) |
---|
80 | { |
---|
81 | long int offset ; |
---|
82 | long int nb=nbCells ; |
---|
83 | ep_lib::MPI_Scan(&nb,&offset,1,EP_LONG,EP_SUM,communicator) ; |
---|
84 | offset=offset-nb ; |
---|
85 | for(int i=0;i<nbCells;i++) targetGlobalId[i]=offset+i ; |
---|
86 | } |
---|
87 | else targetGlobalId.assign(globalId,globalId+nbCells); |
---|
88 | |
---|
89 | for (int i = 0; i < nbCells; i++) |
---|
90 | { |
---|
91 | int offs = i*nVertex; |
---|
92 | Elt elt(boundsLon + offs, boundsLat + offs, nVertex); |
---|
93 | targetElements.push_back(elt); |
---|
94 | targetMesh.push_back(Node(elt.x, cptRadius(elt), &sourceElements.back())); |
---|
95 | cptEltGeom(targetElements[i], Coord(pole[0], pole[1], pole[2])); |
---|
96 | if (area!=NULL) targetElements[i].given_area=area[i] ; |
---|
97 | else targetElements[i].given_area=targetElements[i].area ; |
---|
98 | } |
---|
99 | |
---|
100 | |
---|
101 | } |
---|
102 | |
---|
103 | void Mapper::setSourceValue(const double* val) |
---|
104 | { |
---|
105 | int size=sourceElements.size() ; |
---|
106 | for(int i=0;i<size;++i) sourceElements[i].val=val[i] ; |
---|
107 | } |
---|
108 | |
---|
109 | void Mapper::getTargetValue(double* val) |
---|
110 | { |
---|
111 | int size=targetElements.size() ; |
---|
112 | for(int i=0;i<size;++i) val[i]=targetElements[i].val ; |
---|
113 | } |
---|
114 | |
---|
115 | vector<double> Mapper::computeWeights(int interpOrder, bool renormalize, bool quantity) |
---|
116 | { |
---|
117 | vector<double> timings; |
---|
118 | int mpiSize, mpiRank; |
---|
119 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
120 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
121 | |
---|
122 | this->buildSSTree(sourceMesh, targetMesh); |
---|
123 | |
---|
124 | if (mpiRank == 0 && verbose) cout << "Computing intersections ..." << endl; |
---|
125 | double tic = cputime(); |
---|
126 | computeIntersection(&targetElements[0], targetElements.size()); |
---|
127 | timings.push_back(cputime() - tic); |
---|
128 | |
---|
129 | tic = cputime(); |
---|
130 | if (interpOrder == 2) { |
---|
131 | if (mpiRank == 0 && verbose) cout << "Computing grads ..." << endl; |
---|
132 | buildMeshTopology(); |
---|
133 | computeGrads(); |
---|
134 | } |
---|
135 | timings.push_back(cputime() - tic); |
---|
136 | |
---|
137 | /* Prepare computation of weights */ |
---|
138 | /* compute number of intersections which for the first order case |
---|
139 | corresponds to the number of edges in the remap matrix */ |
---|
140 | int nIntersections = 0; |
---|
141 | for (int j = 0; j < targetElements.size(); j++) |
---|
142 | { |
---|
143 | Elt &elt = targetElements[j]; |
---|
144 | for (list<Polyg*>::iterator it = elt.is.begin(); it != elt.is.end(); it++) |
---|
145 | nIntersections++; |
---|
146 | } |
---|
147 | /* overallocate for NMAX neighbours for each elements */ |
---|
148 | remapMatrix = new double[nIntersections*NMAX]; |
---|
149 | srcAddress = new int[nIntersections*NMAX]; |
---|
150 | srcRank = new int[nIntersections*NMAX]; |
---|
151 | dstAddress = new int[nIntersections*NMAX]; |
---|
152 | sourceWeightId =new long[nIntersections*NMAX]; |
---|
153 | targetWeightId =new long[nIntersections*NMAX]; |
---|
154 | |
---|
155 | |
---|
156 | if (mpiRank == 0 && verbose) cout << "Remapping..." << endl; |
---|
157 | tic = cputime(); |
---|
158 | nWeights = remap(&targetElements[0], targetElements.size(), interpOrder, renormalize, quantity); |
---|
159 | timings.push_back(cputime() - tic); |
---|
160 | |
---|
161 | for (int i = 0; i < targetElements.size(); i++) targetElements[i].delete_intersections(); |
---|
162 | |
---|
163 | return timings; |
---|
164 | } |
---|
165 | |
---|
166 | /** |
---|
167 | @param elements are cells of the target grid that are distributed over CPUs |
---|
168 | indepentently of the distribution of the SS-tree. |
---|
169 | @param nbElements is the size of the elements array. |
---|
170 | @param order is the order of interpolaton (must be 1 or 2). |
---|
171 | */ |
---|
172 | int Mapper::remap(Elt *elements, int nbElements, int order, bool renormalize, bool quantity) |
---|
173 | { |
---|
174 | int mpiSize, mpiRank; |
---|
175 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
176 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
177 | |
---|
178 | /* create list of intersections (super mesh elements) for each rank */ |
---|
179 | multimap<int, Polyg *> *elementList = new multimap<int, Polyg *>[mpiSize]; |
---|
180 | for (int j = 0; j < nbElements; j++) |
---|
181 | { |
---|
182 | Elt& e = elements[j]; |
---|
183 | for (list<Polyg *>::iterator it = e.is.begin(); it != e.is.end(); it++) |
---|
184 | elementList[(*it)->id.rank].insert(pair<int, Polyg *>((*it)->id.ind, *it)); |
---|
185 | } |
---|
186 | |
---|
187 | int *nbSendElement = new int[mpiSize]; |
---|
188 | int **sendElement = new int*[mpiSize]; /* indices of elements required from other rank */ |
---|
189 | double **recvValue = new double*[mpiSize]; |
---|
190 | double **recvArea = new double*[mpiSize]; |
---|
191 | double **recvGivenArea = new double*[mpiSize]; |
---|
192 | Coord **recvGrad = new Coord*[mpiSize]; |
---|
193 | GloId **recvNeighIds = new GloId*[mpiSize]; /* ids of the of the source neighbours which also contribute through gradient */ |
---|
194 | for (int rank = 0; rank < mpiSize; rank++) |
---|
195 | { |
---|
196 | /* get size for allocation */ |
---|
197 | int last = -1; /* compares unequal to any index */ |
---|
198 | int index = -1; /* increased to starting index 0 in first iteration */ |
---|
199 | for (multimap<int, Polyg *>::iterator it = elementList[rank].begin(); it != elementList[rank].end(); ++it) |
---|
200 | { |
---|
201 | if (last != it->first) |
---|
202 | index++; |
---|
203 | (it->second)->id.ind = index; |
---|
204 | last = it->first; |
---|
205 | } |
---|
206 | nbSendElement[rank] = index + 1; |
---|
207 | |
---|
208 | /* if size is non-zero allocate and collect indices of elements on other ranks that we intersect */ |
---|
209 | if (nbSendElement[rank] > 0) |
---|
210 | { |
---|
211 | sendElement[rank] = new int[nbSendElement[rank]]; |
---|
212 | recvValue[rank] = new double[nbSendElement[rank]]; |
---|
213 | recvArea[rank] = new double[nbSendElement[rank]]; |
---|
214 | recvGivenArea[rank] = new double[nbSendElement[rank]]; |
---|
215 | if (order == 2) |
---|
216 | { |
---|
217 | recvNeighIds[rank] = new GloId[nbSendElement[rank]*(NMAX+1)]; |
---|
218 | recvGrad[rank] = new Coord[nbSendElement[rank]*(NMAX+1)]; |
---|
219 | } |
---|
220 | else |
---|
221 | recvNeighIds[rank] = new GloId[nbSendElement[rank]]; |
---|
222 | |
---|
223 | last = -1; |
---|
224 | index = -1; |
---|
225 | for (multimap<int, Polyg *>::iterator it = elementList[rank].begin(); it != elementList[rank].end(); ++it) |
---|
226 | { |
---|
227 | if (last != it->first) |
---|
228 | index++; |
---|
229 | sendElement[rank][index] = it->first; |
---|
230 | last = it->first; |
---|
231 | } |
---|
232 | } |
---|
233 | } |
---|
234 | |
---|
235 | /* communicate sizes of source elements to be sent (index lists and later values and gradients) */ |
---|
236 | int *nbRecvElement = new int[mpiSize]; |
---|
237 | ep_lib::MPI_Alltoall(nbSendElement, 1, EP_INT, nbRecvElement, 1, EP_INT, communicator); |
---|
238 | |
---|
239 | /* communicate indices of source elements on other ranks whoes value and gradient we need (since intersection) */ |
---|
240 | int nbSendRequest = 0; |
---|
241 | int nbRecvRequest = 0; |
---|
242 | int **recvElement = new int*[mpiSize]; |
---|
243 | double **sendValue = new double*[mpiSize]; |
---|
244 | double **sendArea = new double*[mpiSize]; |
---|
245 | double **sendGivenArea = new double*[mpiSize]; |
---|
246 | Coord **sendGrad = new Coord*[mpiSize]; |
---|
247 | GloId **sendNeighIds = new GloId*[mpiSize]; |
---|
248 | ep_lib::MPI_Request *sendRequest = new ep_lib::MPI_Request[5*mpiSize]; |
---|
249 | ep_lib::MPI_Request *recvRequest = new ep_lib::MPI_Request[5*mpiSize]; |
---|
250 | for (int rank = 0; rank < mpiSize; rank++) |
---|
251 | { |
---|
252 | if (nbSendElement[rank] > 0) |
---|
253 | { |
---|
254 | ep_lib::MPI_Issend(sendElement[rank], nbSendElement[rank], EP_INT, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
255 | nbSendRequest++; |
---|
256 | } |
---|
257 | |
---|
258 | if (nbRecvElement[rank] > 0) |
---|
259 | { |
---|
260 | recvElement[rank] = new int[nbRecvElement[rank]]; |
---|
261 | sendValue[rank] = new double[nbRecvElement[rank]]; |
---|
262 | sendArea[rank] = new double[nbRecvElement[rank]]; |
---|
263 | sendGivenArea[rank] = new double[nbRecvElement[rank]]; |
---|
264 | if (order == 2) |
---|
265 | { |
---|
266 | sendNeighIds[rank] = new GloId[nbRecvElement[rank]*(NMAX+1)]; |
---|
267 | sendGrad[rank] = new Coord[nbRecvElement[rank]*(NMAX+1)]; |
---|
268 | } |
---|
269 | else |
---|
270 | { |
---|
271 | sendNeighIds[rank] = new GloId[nbRecvElement[rank]]; |
---|
272 | } |
---|
273 | ep_lib::MPI_Irecv(recvElement[rank], nbRecvElement[rank], EP_INT, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
274 | nbRecvRequest++; |
---|
275 | } |
---|
276 | } |
---|
277 | ep_lib::MPI_Status *status = new ep_lib::MPI_Status[5*mpiSize]; |
---|
278 | |
---|
279 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
280 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
281 | |
---|
282 | /* for all indices that have been received from requesting ranks: pack values and gradients, then send */ |
---|
283 | nbSendRequest = 0; |
---|
284 | nbRecvRequest = 0; |
---|
285 | for (int rank = 0; rank < mpiSize; rank++) |
---|
286 | { |
---|
287 | if (nbRecvElement[rank] > 0) |
---|
288 | { |
---|
289 | int jj = 0; // jj == j if no weight writing |
---|
290 | for (int j = 0; j < nbRecvElement[rank]; j++) |
---|
291 | { |
---|
292 | sendValue[rank][j] = sstree.localElements[recvElement[rank][j]].val; |
---|
293 | sendArea[rank][j] = sstree.localElements[recvElement[rank][j]].area; |
---|
294 | sendGivenArea[rank][j] = sstree.localElements[recvElement[rank][j]].given_area; |
---|
295 | if (order == 2) |
---|
296 | { |
---|
297 | sendGrad[rank][jj] = sstree.localElements[recvElement[rank][j]].grad; |
---|
298 | // cout<<"grad "<<jj<<" "<<recvElement[rank][j]<<" "<<sendGrad[rank][jj]<<" "<<sstree.localElements[recvElement[rank][j]].grad<<endl ; |
---|
299 | sendNeighIds[rank][jj] = sstree.localElements[recvElement[rank][j]].src_id; |
---|
300 | jj++; |
---|
301 | for (int i = 0; i < NMAX; i++) |
---|
302 | { |
---|
303 | sendGrad[rank][jj] = sstree.localElements[recvElement[rank][j]].gradNeigh[i]; |
---|
304 | // cout<<"grad "<<jj<<" "<<sendGrad[rank][jj]<<" "<<sstree.localElements[recvElement[rank][j]].grad<<endl ; |
---|
305 | sendNeighIds[rank][jj] = sstree.localElements[recvElement[rank][j]].neighId[i]; |
---|
306 | jj++; |
---|
307 | } |
---|
308 | } |
---|
309 | else |
---|
310 | sendNeighIds[rank][j] = sstree.localElements[recvElement[rank][j]].src_id; |
---|
311 | } |
---|
312 | ep_lib::MPI_Issend(sendValue[rank], nbRecvElement[rank], EP_DOUBLE, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
313 | nbSendRequest++; |
---|
314 | ep_lib::MPI_Issend(sendArea[rank], nbRecvElement[rank], EP_DOUBLE, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
315 | nbSendRequest++; |
---|
316 | ep_lib::MPI_Issend(sendGivenArea[rank], nbRecvElement[rank], EP_DOUBLE, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
317 | nbSendRequest++; |
---|
318 | if (order == 2) |
---|
319 | { |
---|
320 | ep_lib::MPI_Issend(sendGrad[rank], 3*nbRecvElement[rank]*(NMAX+1), EP_DOUBLE, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
321 | nbSendRequest++; |
---|
322 | ep_lib::MPI_Issend(sendNeighIds[rank], 4*nbRecvElement[rank]*(NMAX+1), EP_INT, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
323 | //ym --> attention taille GloId |
---|
324 | nbSendRequest++; |
---|
325 | } |
---|
326 | else |
---|
327 | { |
---|
328 | ep_lib::MPI_Issend(sendNeighIds[rank], 4*nbRecvElement[rank], EP_INT, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
329 | //ym --> attention taille GloId |
---|
330 | nbSendRequest++; |
---|
331 | } |
---|
332 | } |
---|
333 | if (nbSendElement[rank] > 0) |
---|
334 | { |
---|
335 | ep_lib::MPI_Irecv(recvValue[rank], nbSendElement[rank], EP_DOUBLE, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
336 | nbRecvRequest++; |
---|
337 | ep_lib::MPI_Irecv(recvArea[rank], nbSendElement[rank], EP_DOUBLE, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
338 | nbRecvRequest++; |
---|
339 | ep_lib::MPI_Irecv(recvGivenArea[rank], nbSendElement[rank], EP_DOUBLE, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
340 | nbRecvRequest++; |
---|
341 | if (order == 2) |
---|
342 | { |
---|
343 | ep_lib::MPI_Irecv(recvGrad[rank], 3*nbSendElement[rank]*(NMAX+1), |
---|
344 | EP_DOUBLE, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
345 | nbRecvRequest++; |
---|
346 | ep_lib::MPI_Irecv(recvNeighIds[rank], 4*nbSendElement[rank]*(NMAX+1), EP_INT, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
347 | //ym --> attention taille GloId |
---|
348 | nbRecvRequest++; |
---|
349 | } |
---|
350 | else |
---|
351 | { |
---|
352 | ep_lib::MPI_Irecv(recvNeighIds[rank], 4*nbSendElement[rank], EP_INT, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
353 | //ym --> attention taille GloId |
---|
354 | nbRecvRequest++; |
---|
355 | } |
---|
356 | } |
---|
357 | } |
---|
358 | |
---|
359 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
360 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
361 | |
---|
362 | |
---|
363 | /* now that all values and gradients are available use them to computed interpolated values on target |
---|
364 | and also to compute weights */ |
---|
365 | int i = 0; |
---|
366 | for (int j = 0; j < nbElements; j++) |
---|
367 | { |
---|
368 | Elt& e = elements[j]; |
---|
369 | |
---|
370 | /* since for the 2nd order case source grid elements can contribute to a destination grid element over several "paths" |
---|
371 | (step1: gradient is computed using neighbours on same grid, step2: intersection uses several elements on other grid) |
---|
372 | accumulate them so that there is only one final weight between two elements */ |
---|
373 | map<GloId,double> wgt_map; |
---|
374 | |
---|
375 | /* for destination element `e` loop over all intersetions/the corresponding source elements */ |
---|
376 | for (list<Polyg *>::iterator it = e.is.begin(); it != e.is.end(); it++) |
---|
377 | { |
---|
378 | /* it is the intersection element, so it->x and it->area are barycentre and area of intersection element (super mesh) |
---|
379 | but it->id is id of the source element that it intersects */ |
---|
380 | int n1 = (*it)->id.ind; |
---|
381 | int rank = (*it)->id.rank; |
---|
382 | double fk = recvValue[rank][n1]; |
---|
383 | double srcArea = recvArea[rank][n1]; |
---|
384 | double srcGivenArea = recvGivenArea[rank][n1]; |
---|
385 | double w = (*it)->area; |
---|
386 | if (quantity) w/=srcArea ; |
---|
387 | else w=w*srcGivenArea/srcArea*e.area/e.given_area ; |
---|
388 | |
---|
389 | /* first order: src value times weight (weight = supermesh area), later divide by target area */ |
---|
390 | int kk = (order == 2) ? n1 * (NMAX + 1) : n1; |
---|
391 | GloId neighID = recvNeighIds[rank][kk]; |
---|
392 | wgt_map[neighID] += w; |
---|
393 | |
---|
394 | if (order == 2) |
---|
395 | { |
---|
396 | for (int k = 0; k < NMAX+1; k++) |
---|
397 | { |
---|
398 | int kk = n1 * (NMAX + 1) + k; |
---|
399 | GloId neighID = recvNeighIds[rank][kk]; |
---|
400 | if (neighID.ind != -1) wgt_map[neighID] += w * scalarprod(recvGrad[rank][kk], (*it)->x); |
---|
401 | } |
---|
402 | |
---|
403 | } |
---|
404 | } |
---|
405 | |
---|
406 | double renorm=0; |
---|
407 | if (renormalize) |
---|
408 | { |
---|
409 | if (quantity) for (map<GloId,double>::iterator it = wgt_map.begin(); it != wgt_map.end(); it++) renorm+=it->second ; |
---|
410 | else for (map<GloId,double>::iterator it = wgt_map.begin(); it != wgt_map.end(); it++) renorm+=it->second / e.area; |
---|
411 | } |
---|
412 | else renorm=1. ; |
---|
413 | |
---|
414 | for (map<GloId,double>::iterator it = wgt_map.begin(); it != wgt_map.end(); it++) |
---|
415 | { |
---|
416 | if (quantity) this->remapMatrix[i] = (it->second ) / renorm; |
---|
417 | else this->remapMatrix[i] = (it->second / e.area) / renorm; |
---|
418 | this->srcAddress[i] = it->first.ind; |
---|
419 | this->srcRank[i] = it->first.rank; |
---|
420 | this->dstAddress[i] = j; |
---|
421 | this->sourceWeightId[i]= it->first.globalId ; |
---|
422 | this->targetWeightId[i]= targetGlobalId[j] ; |
---|
423 | i++; |
---|
424 | } |
---|
425 | } |
---|
426 | |
---|
427 | /* free all memory allocated in this function */ |
---|
428 | for (int rank = 0; rank < mpiSize; rank++) |
---|
429 | { |
---|
430 | if (nbSendElement[rank] > 0) |
---|
431 | { |
---|
432 | delete[] sendElement[rank]; |
---|
433 | delete[] recvValue[rank]; |
---|
434 | delete[] recvArea[rank]; |
---|
435 | delete[] recvGivenArea[rank]; |
---|
436 | if (order == 2) |
---|
437 | { |
---|
438 | delete[] recvGrad[rank]; |
---|
439 | } |
---|
440 | delete[] recvNeighIds[rank]; |
---|
441 | } |
---|
442 | if (nbRecvElement[rank] > 0) |
---|
443 | { |
---|
444 | delete[] recvElement[rank]; |
---|
445 | delete[] sendValue[rank]; |
---|
446 | delete[] sendArea[rank]; |
---|
447 | delete[] sendGivenArea[rank]; |
---|
448 | if (order == 2) |
---|
449 | delete[] sendGrad[rank]; |
---|
450 | delete[] sendNeighIds[rank]; |
---|
451 | } |
---|
452 | } |
---|
453 | delete[] status; |
---|
454 | delete[] sendRequest; |
---|
455 | delete[] recvRequest; |
---|
456 | delete[] elementList; |
---|
457 | delete[] nbSendElement; |
---|
458 | delete[] nbRecvElement; |
---|
459 | delete[] sendElement; |
---|
460 | delete[] recvElement; |
---|
461 | delete[] sendValue; |
---|
462 | delete[] recvValue; |
---|
463 | delete[] sendGrad; |
---|
464 | delete[] recvGrad; |
---|
465 | delete[] sendNeighIds; |
---|
466 | delete[] recvNeighIds; |
---|
467 | return i; |
---|
468 | } |
---|
469 | |
---|
470 | void Mapper::computeGrads() |
---|
471 | { |
---|
472 | /* array of pointers to collect local elements and elements received from other cpu */ |
---|
473 | vector<Elt*> globalElements(sstree.nbLocalElements + nbNeighbourElements); |
---|
474 | int index = 0; |
---|
475 | for (int i = 0; i < sstree.nbLocalElements; i++, index++) |
---|
476 | globalElements[index] = &(sstree.localElements[i]); |
---|
477 | for (int i = 0; i < nbNeighbourElements; i++, index++) |
---|
478 | globalElements[index] = &neighbourElements[i]; |
---|
479 | |
---|
480 | update_baryc(sstree.localElements, sstree.nbLocalElements); |
---|
481 | computeGradients(&globalElements[0], sstree.nbLocalElements); |
---|
482 | } |
---|
483 | |
---|
484 | /** for each element of the source grid, finds all the neighbouring elements that share an edge |
---|
485 | (filling array neighbourElements). This is used later to compute gradients */ |
---|
486 | void Mapper::buildMeshTopology() |
---|
487 | { |
---|
488 | int mpiSize, mpiRank; |
---|
489 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
490 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
491 | |
---|
492 | vector<Node> *routingList = new vector<Node>[mpiSize]; |
---|
493 | vector<vector<int> > routes(sstree.localTree.leafs.size()); |
---|
494 | |
---|
495 | sstree.routeIntersections(routes, sstree.localTree.leafs); |
---|
496 | |
---|
497 | for (int i = 0; i < routes.size(); ++i) |
---|
498 | for (int k = 0; k < routes[i].size(); ++k) |
---|
499 | routingList[routes[i][k]].push_back(sstree.localTree.leafs[i]); |
---|
500 | routingList[mpiRank].clear(); |
---|
501 | |
---|
502 | |
---|
503 | CMPIRouting mpiRoute(communicator); |
---|
504 | mpiRoute.init(routes); |
---|
505 | int nRecv = mpiRoute.getTotalSourceElement(); |
---|
506 | // cout << mpiRank << " NRECV " << nRecv << "(" << routes.size() << ")"<< endl; |
---|
507 | |
---|
508 | int *nbSendNode = new int[mpiSize]; |
---|
509 | int *nbRecvNode = new int[mpiSize]; |
---|
510 | int *sendMessageSize = new int[mpiSize]; |
---|
511 | int *recvMessageSize = new int[mpiSize]; |
---|
512 | |
---|
513 | for (int rank = 0; rank < mpiSize; rank++) |
---|
514 | { |
---|
515 | nbSendNode[rank] = routingList[rank].size(); |
---|
516 | sendMessageSize[rank] = 0; |
---|
517 | for (size_t j = 0; j < routingList[rank].size(); j++) |
---|
518 | { |
---|
519 | Elt *elt = (Elt *) (routingList[rank][j].data); |
---|
520 | sendMessageSize[rank] += packedPolygonSize(*elt); |
---|
521 | } |
---|
522 | } |
---|
523 | |
---|
524 | ep_lib::MPI_Alltoall(nbSendNode, 1, EP_INT, nbRecvNode, 1, EP_INT, communicator); |
---|
525 | ep_lib::MPI_Alltoall(sendMessageSize, 1, EP_INT, recvMessageSize, 1, EP_INT, communicator); |
---|
526 | |
---|
527 | char **sendBuffer = new char*[mpiSize]; |
---|
528 | char **recvBuffer = new char*[mpiSize]; |
---|
529 | int *pos = new int[mpiSize]; |
---|
530 | |
---|
531 | for (int rank = 0; rank < mpiSize; rank++) |
---|
532 | { |
---|
533 | if (nbSendNode[rank] > 0) sendBuffer[rank] = new char[sendMessageSize[rank]]; |
---|
534 | if (nbRecvNode[rank] > 0) recvBuffer[rank] = new char[recvMessageSize[rank]]; |
---|
535 | } |
---|
536 | |
---|
537 | for (int rank = 0; rank < mpiSize; rank++) |
---|
538 | { |
---|
539 | pos[rank] = 0; |
---|
540 | for (size_t j = 0; j < routingList[rank].size(); j++) |
---|
541 | { |
---|
542 | Elt *elt = (Elt *) (routingList[rank][j].data); |
---|
543 | packPolygon(*elt, sendBuffer[rank], pos[rank]); |
---|
544 | } |
---|
545 | } |
---|
546 | delete [] routingList; |
---|
547 | |
---|
548 | |
---|
549 | int nbSendRequest = 0; |
---|
550 | int nbRecvRequest = 0; |
---|
551 | ep_lib::MPI_Request *sendRequest = new ep_lib::MPI_Request[mpiSize]; |
---|
552 | ep_lib::MPI_Request *recvRequest = new ep_lib::MPI_Request[mpiSize]; |
---|
553 | ep_lib::MPI_Status *status = new ep_lib::MPI_Status[mpiSize]; |
---|
554 | |
---|
555 | for (int rank = 0; rank < mpiSize; rank++) |
---|
556 | { |
---|
557 | if (nbSendNode[rank] > 0) |
---|
558 | { |
---|
559 | ep_lib::MPI_Issend(sendBuffer[rank], sendMessageSize[rank], EP_CHAR, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
560 | nbSendRequest++; |
---|
561 | } |
---|
562 | if (nbRecvNode[rank] > 0) |
---|
563 | { |
---|
564 | ep_lib::MPI_Irecv(recvBuffer[rank], recvMessageSize[rank], EP_CHAR, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
565 | nbRecvRequest++; |
---|
566 | } |
---|
567 | } |
---|
568 | |
---|
569 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
570 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
571 | |
---|
572 | for (int rank = 0; rank < mpiSize; rank++) |
---|
573 | if (nbSendNode[rank] > 0) delete [] sendBuffer[rank]; |
---|
574 | delete [] sendBuffer; |
---|
575 | |
---|
576 | char **sendBuffer2 = new char*[mpiSize]; |
---|
577 | char **recvBuffer2 = new char*[mpiSize]; |
---|
578 | |
---|
579 | for (int rank = 0; rank < mpiSize; rank++) |
---|
580 | { |
---|
581 | nbSendNode[rank] = 0; |
---|
582 | sendMessageSize[rank] = 0; |
---|
583 | |
---|
584 | if (nbRecvNode[rank] > 0) |
---|
585 | { |
---|
586 | set<NodePtr> neighbourList; |
---|
587 | pos[rank] = 0; |
---|
588 | for (int j = 0; j < nbRecvNode[rank]; j++) |
---|
589 | { |
---|
590 | Elt elt; |
---|
591 | unpackPolygon(elt, recvBuffer[rank], pos[rank]); |
---|
592 | Node node(elt.x, cptRadius(elt), &elt); |
---|
593 | findNeighbour(sstree.localTree.root, &node, neighbourList); |
---|
594 | } |
---|
595 | nbSendNode[rank] = neighbourList.size(); |
---|
596 | for (set<NodePtr>::iterator it = neighbourList.begin(); it != neighbourList.end(); it++) |
---|
597 | { |
---|
598 | Elt *elt = (Elt *) ((*it)->data); |
---|
599 | sendMessageSize[rank] += packedPolygonSize(*elt); |
---|
600 | } |
---|
601 | |
---|
602 | sendBuffer2[rank] = new char[sendMessageSize[rank]]; |
---|
603 | pos[rank] = 0; |
---|
604 | |
---|
605 | for (set<NodePtr>::iterator it = neighbourList.begin(); it != neighbourList.end(); it++) |
---|
606 | { |
---|
607 | Elt *elt = (Elt *) ((*it)->data); |
---|
608 | packPolygon(*elt, sendBuffer2[rank], pos[rank]); |
---|
609 | } |
---|
610 | } |
---|
611 | } |
---|
612 | for (int rank = 0; rank < mpiSize; rank++) |
---|
613 | if (nbRecvNode[rank] > 0) delete [] recvBuffer[rank]; |
---|
614 | delete [] recvBuffer; |
---|
615 | |
---|
616 | |
---|
617 | ep_lib::MPI_Barrier(communicator); |
---|
618 | ep_lib::MPI_Alltoall(nbSendNode, 1, EP_INT, nbRecvNode, 1, EP_INT, communicator); |
---|
619 | ep_lib::MPI_Alltoall(sendMessageSize, 1, EP_INT, recvMessageSize, 1, EP_INT, communicator); |
---|
620 | |
---|
621 | for (int rank = 0; rank < mpiSize; rank++) |
---|
622 | if (nbRecvNode[rank] > 0) recvBuffer2[rank] = new char[recvMessageSize[rank]]; |
---|
623 | |
---|
624 | nbSendRequest = 0; |
---|
625 | nbRecvRequest = 0; |
---|
626 | |
---|
627 | for (int rank = 0; rank < mpiSize; rank++) |
---|
628 | { |
---|
629 | if (nbSendNode[rank] > 0) |
---|
630 | { |
---|
631 | ep_lib::MPI_Issend(sendBuffer2[rank], sendMessageSize[rank], EP_CHAR, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
632 | nbSendRequest++; |
---|
633 | } |
---|
634 | if (nbRecvNode[rank] > 0) |
---|
635 | { |
---|
636 | ep_lib::MPI_Irecv(recvBuffer2[rank], recvMessageSize[rank], EP_CHAR, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
637 | nbRecvRequest++; |
---|
638 | } |
---|
639 | } |
---|
640 | |
---|
641 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
642 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
643 | |
---|
644 | int nbNeighbourNodes = 0; |
---|
645 | for (int rank = 0; rank < mpiSize; rank++) |
---|
646 | nbNeighbourNodes += nbRecvNode[rank]; |
---|
647 | |
---|
648 | neighbourElements = new Elt[nbNeighbourNodes]; |
---|
649 | nbNeighbourElements = nbNeighbourNodes; |
---|
650 | |
---|
651 | int index = 0; |
---|
652 | for (int rank = 0; rank < mpiSize; rank++) |
---|
653 | { |
---|
654 | pos[rank] = 0; |
---|
655 | for (int j = 0; j < nbRecvNode[rank]; j++) |
---|
656 | { |
---|
657 | unpackPolygon(neighbourElements[index], recvBuffer2[rank], pos[rank]); |
---|
658 | neighbourElements[index].id.ind = sstree.localTree.leafs.size() + index; |
---|
659 | index++; |
---|
660 | } |
---|
661 | } |
---|
662 | for (int rank = 0; rank < mpiSize; rank++) |
---|
663 | { |
---|
664 | if (nbRecvNode[rank] > 0) delete [] recvBuffer2[rank]; |
---|
665 | if (nbSendNode[rank] > 0) delete [] sendBuffer2[rank]; |
---|
666 | } |
---|
667 | delete [] recvBuffer2; |
---|
668 | delete [] sendBuffer2; |
---|
669 | delete [] sendMessageSize; |
---|
670 | delete [] recvMessageSize; |
---|
671 | delete [] nbSendNode; |
---|
672 | delete [] nbRecvNode; |
---|
673 | delete [] sendRequest; |
---|
674 | delete [] recvRequest; |
---|
675 | delete [] status; |
---|
676 | delete [] pos; |
---|
677 | |
---|
678 | /* re-compute on received elements to avoid having to send this information */ |
---|
679 | neighbourNodes.resize(nbNeighbourNodes); |
---|
680 | setCirclesAndLinks(neighbourElements, neighbourNodes); |
---|
681 | cptAllEltsGeom(neighbourElements, nbNeighbourNodes, srcGrid.pole); |
---|
682 | |
---|
683 | /* the local SS tree must include nodes from other cpus if they are potential |
---|
684 | intersector of a local node */ |
---|
685 | sstree.localTree.insertNodes(neighbourNodes); |
---|
686 | |
---|
687 | /* for every local element, |
---|
688 | use the SS-tree to find all elements (including neighbourElements) |
---|
689 | who are potential neighbours because their circles intersect, |
---|
690 | then check all canditates for common edges to build up connectivity information |
---|
691 | */ |
---|
692 | for (int j = 0; j < sstree.localTree.leafs.size(); j++) |
---|
693 | { |
---|
694 | Node& node = sstree.localTree.leafs[j]; |
---|
695 | |
---|
696 | /* find all leafs whoes circles that intersect node's circle and save into node->intersectors */ |
---|
697 | node.search(sstree.localTree.root); |
---|
698 | |
---|
699 | Elt *elt = (Elt *)(node.data); |
---|
700 | |
---|
701 | for (int i = 0; i < elt->n; i++) elt->neighbour[i] = NOT_FOUND; |
---|
702 | |
---|
703 | /* for element `elt` loop through all nodes in the SS-tree |
---|
704 | whoes circles intersect with the circle around `elt` (the SS intersectors) |
---|
705 | and check if they are neighbours in the sense that the two elements share an edge. |
---|
706 | If they do, save this information for elt */ |
---|
707 | for (list<NodePtr>::iterator it = (node.intersectors).begin(); it != (node.intersectors).end(); ++it) |
---|
708 | { |
---|
709 | Elt *elt2 = (Elt *)((*it)->data); |
---|
710 | set_neighbour(*elt, *elt2); |
---|
711 | } |
---|
712 | |
---|
713 | /* |
---|
714 | for (int i = 0; i < elt->n; i++) |
---|
715 | { |
---|
716 | if (elt->neighbour[i] == NOT_FOUND) |
---|
717 | error_exit("neighbour not found"); |
---|
718 | } |
---|
719 | */ |
---|
720 | } |
---|
721 | } |
---|
722 | |
---|
723 | /** @param elements are the target grid elements */ |
---|
724 | void Mapper::computeIntersection(Elt *elements, int nbElements) |
---|
725 | { |
---|
726 | int mpiSize, mpiRank; |
---|
727 | ep_lib::MPI_Comm_size(communicator, &mpiSize); |
---|
728 | ep_lib::MPI_Comm_rank(communicator, &mpiRank); |
---|
729 | |
---|
730 | ep_lib::MPI_Barrier(communicator); |
---|
731 | |
---|
732 | vector<Node> *routingList = new vector<Node>[mpiSize]; |
---|
733 | |
---|
734 | vector<Node> routeNodes; routeNodes.reserve(nbElements); |
---|
735 | for (int j = 0; j < nbElements; j++) |
---|
736 | { |
---|
737 | elements[j].id.ind = j; |
---|
738 | elements[j].id.rank = mpiRank; |
---|
739 | routeNodes.push_back(Node(elements[j].x, cptRadius(elements[j]), &elements[j])); |
---|
740 | } |
---|
741 | |
---|
742 | vector<vector<int> > routes(routeNodes.size()); |
---|
743 | sstree.routeIntersections(routes, routeNodes); |
---|
744 | for (int i = 0; i < routes.size(); ++i) |
---|
745 | for (int k = 0; k < routes[i].size(); ++k) |
---|
746 | routingList[routes[i][k]].push_back(routeNodes[i]); |
---|
747 | |
---|
748 | if (verbose >= 2) |
---|
749 | { |
---|
750 | cout << " --> rank " << mpiRank << " nbElements " << nbElements << " : "; |
---|
751 | for (int rank = 0; rank < mpiSize; rank++) |
---|
752 | cout << routingList[rank].size() << " "; |
---|
753 | cout << endl; |
---|
754 | } |
---|
755 | ep_lib::MPI_Barrier(communicator); |
---|
756 | |
---|
757 | int *nbSendNode = new int[mpiSize]; |
---|
758 | int *nbRecvNode = new int[mpiSize]; |
---|
759 | int *sentMessageSize = new int[mpiSize]; |
---|
760 | int *recvMessageSize = new int[mpiSize]; |
---|
761 | |
---|
762 | for (int rank = 0; rank < mpiSize; rank++) |
---|
763 | { |
---|
764 | nbSendNode[rank] = routingList[rank].size(); |
---|
765 | sentMessageSize[rank] = 0; |
---|
766 | for (size_t j = 0; j < routingList[rank].size(); j++) |
---|
767 | { |
---|
768 | Elt *elt = (Elt *) (routingList[rank][j].data); |
---|
769 | sentMessageSize[rank] += packedPolygonSize(*elt); |
---|
770 | } |
---|
771 | } |
---|
772 | |
---|
773 | ep_lib::MPI_Alltoall(nbSendNode, 1, EP_INT, nbRecvNode, 1, EP_INT, communicator); |
---|
774 | ep_lib::MPI_Alltoall(sentMessageSize, 1, EP_INT, recvMessageSize, 1, EP_INT, communicator); |
---|
775 | |
---|
776 | int total = 0; |
---|
777 | |
---|
778 | for (int rank = 0; rank < mpiSize; rank++) |
---|
779 | { |
---|
780 | total = total + nbRecvNode[rank]; |
---|
781 | } |
---|
782 | |
---|
783 | if (verbose >= 2) cout << "---> rank " << mpiRank << " : compute intersection : total received nodes " << total << endl; |
---|
784 | char **sendBuffer = new char*[mpiSize]; |
---|
785 | char **recvBuffer = new char*[mpiSize]; |
---|
786 | int *pos = new int[mpiSize]; |
---|
787 | |
---|
788 | for (int rank = 0; rank < mpiSize; rank++) |
---|
789 | { |
---|
790 | if (nbSendNode[rank] > 0) sendBuffer[rank] = new char[sentMessageSize[rank]]; |
---|
791 | if (nbRecvNode[rank] > 0) recvBuffer[rank] = new char[recvMessageSize[rank]]; |
---|
792 | } |
---|
793 | |
---|
794 | for (int rank = 0; rank < mpiSize; rank++) |
---|
795 | { |
---|
796 | pos[rank] = 0; |
---|
797 | for (size_t j = 0; j < routingList[rank].size(); j++) |
---|
798 | { |
---|
799 | Elt* elt = (Elt *) (routingList[rank][j].data); |
---|
800 | packPolygon(*elt, sendBuffer[rank], pos[rank]); |
---|
801 | } |
---|
802 | } |
---|
803 | delete [] routingList; |
---|
804 | |
---|
805 | int nbSendRequest = 0; |
---|
806 | int nbRecvRequest = 0; |
---|
807 | ep_lib::MPI_Request *sendRequest = new ep_lib::MPI_Request[mpiSize]; |
---|
808 | ep_lib::MPI_Request *recvRequest = new ep_lib::MPI_Request[mpiSize]; |
---|
809 | ep_lib::MPI_Status *status = new ep_lib::MPI_Status[mpiSize]; |
---|
810 | |
---|
811 | for (int rank = 0; rank < mpiSize; rank++) |
---|
812 | { |
---|
813 | if (nbSendNode[rank] > 0) |
---|
814 | { |
---|
815 | ep_lib::MPI_Issend(sendBuffer[rank], sentMessageSize[rank], EP_CHAR, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
816 | nbSendRequest++; |
---|
817 | } |
---|
818 | if (nbRecvNode[rank] > 0) |
---|
819 | { |
---|
820 | ep_lib::MPI_Irecv(recvBuffer[rank], recvMessageSize[rank], EP_CHAR, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
821 | nbRecvRequest++; |
---|
822 | } |
---|
823 | } |
---|
824 | |
---|
825 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
826 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
827 | char **sendBuffer2 = new char*[mpiSize]; |
---|
828 | char **recvBuffer2 = new char*[mpiSize]; |
---|
829 | |
---|
830 | double tic = cputime(); |
---|
831 | for (int rank = 0; rank < mpiSize; rank++) |
---|
832 | { |
---|
833 | sentMessageSize[rank] = 0; |
---|
834 | |
---|
835 | if (nbRecvNode[rank] > 0) |
---|
836 | { |
---|
837 | Elt *recvElt = new Elt[nbRecvNode[rank]]; |
---|
838 | pos[rank] = 0; |
---|
839 | for (int j = 0; j < nbRecvNode[rank]; j++) |
---|
840 | { |
---|
841 | unpackPolygon(recvElt[j], recvBuffer[rank], pos[rank]); |
---|
842 | cptEltGeom(recvElt[j], tgtGrid.pole); |
---|
843 | Node recvNode(recvElt[j].x, cptRadius(recvElt[j]), &recvElt[j]); |
---|
844 | recvNode.search(sstree.localTree.root); |
---|
845 | /* for a node holding an element of the target, loop throught candidates for intersecting source */ |
---|
846 | for (list<NodePtr>::iterator it = (recvNode.intersectors).begin(); it != (recvNode.intersectors).end(); ++it) |
---|
847 | { |
---|
848 | Elt *elt2 = (Elt *) ((*it)->data); |
---|
849 | /* recvElt is target, elt2 is source */ |
---|
850 | // intersect(&recvElt[j], elt2); |
---|
851 | intersect_ym(&recvElt[j], elt2); |
---|
852 | } |
---|
853 | |
---|
854 | if (recvElt[j].is.size() > 0) sentMessageSize[rank] += packIntersectionSize(recvElt[j]); |
---|
855 | |
---|
856 | // here recvNode goes out of scope |
---|
857 | } |
---|
858 | |
---|
859 | if (sentMessageSize[rank] > 0) |
---|
860 | { |
---|
861 | sentMessageSize[rank] += sizeof(int); |
---|
862 | sendBuffer2[rank] = new char[sentMessageSize[rank]]; |
---|
863 | *((int *) sendBuffer2[rank]) = 0; |
---|
864 | pos[rank] = sizeof(int); |
---|
865 | for (int j = 0; j < nbRecvNode[rank]; j++) |
---|
866 | { |
---|
867 | packIntersection(recvElt[j], sendBuffer2[rank], pos[rank]); |
---|
868 | //FIXME should be deleted: recvElt[j].delete_intersections(); // intersection areas have been packed to buffer and won't be used any more |
---|
869 | } |
---|
870 | } |
---|
871 | delete [] recvElt; |
---|
872 | |
---|
873 | } |
---|
874 | } |
---|
875 | delete [] pos; |
---|
876 | |
---|
877 | for (int rank = 0; rank < mpiSize; rank++) |
---|
878 | { |
---|
879 | if (nbSendNode[rank] > 0) delete [] sendBuffer[rank]; |
---|
880 | if (nbRecvNode[rank] > 0) delete [] recvBuffer[rank]; |
---|
881 | nbSendNode[rank] = 0; |
---|
882 | } |
---|
883 | |
---|
884 | if (verbose >= 2) cout << "Rank " << mpiRank << " Compute (internal) intersection " << cputime() - tic << " s" << endl; |
---|
885 | ep_lib::MPI_Alltoall(sentMessageSize, 1, EP_INT, recvMessageSize, 1, EP_INT, communicator); |
---|
886 | |
---|
887 | for (int rank = 0; rank < mpiSize; rank++) |
---|
888 | if (recvMessageSize[rank] > 0) |
---|
889 | recvBuffer2[rank] = new char[recvMessageSize[rank]]; |
---|
890 | |
---|
891 | nbSendRequest = 0; |
---|
892 | nbRecvRequest = 0; |
---|
893 | |
---|
894 | for (int rank = 0; rank < mpiSize; rank++) |
---|
895 | { |
---|
896 | if (sentMessageSize[rank] > 0) |
---|
897 | { |
---|
898 | ep_lib::MPI_Issend(sendBuffer2[rank], sentMessageSize[rank], EP_CHAR, rank, 0, communicator, &sendRequest[nbSendRequest]); |
---|
899 | nbSendRequest++; |
---|
900 | } |
---|
901 | if (recvMessageSize[rank] > 0) |
---|
902 | { |
---|
903 | ep_lib::MPI_Irecv(recvBuffer2[rank], recvMessageSize[rank], EP_CHAR, rank, 0, communicator, &recvRequest[nbRecvRequest]); |
---|
904 | nbRecvRequest++; |
---|
905 | } |
---|
906 | } |
---|
907 | |
---|
908 | ep_lib::MPI_Waitall(nbRecvRequest, recvRequest, status); |
---|
909 | ep_lib::MPI_Waitall(nbSendRequest, sendRequest, status); |
---|
910 | |
---|
911 | delete [] sendRequest; |
---|
912 | delete [] recvRequest; |
---|
913 | delete [] status; |
---|
914 | for (int rank = 0; rank < mpiSize; rank++) |
---|
915 | { |
---|
916 | if (nbRecvNode[rank] > 0) |
---|
917 | { |
---|
918 | if (sentMessageSize[rank] > 0) |
---|
919 | delete [] sendBuffer2[rank]; |
---|
920 | } |
---|
921 | |
---|
922 | if (recvMessageSize[rank] > 0) |
---|
923 | { |
---|
924 | unpackIntersection(elements, recvBuffer2[rank]); |
---|
925 | delete [] recvBuffer2[rank]; |
---|
926 | } |
---|
927 | } |
---|
928 | delete [] sendBuffer2; |
---|
929 | delete [] recvBuffer2; |
---|
930 | delete [] sendBuffer; |
---|
931 | delete [] recvBuffer; |
---|
932 | |
---|
933 | delete [] nbSendNode; |
---|
934 | delete [] nbRecvNode; |
---|
935 | delete [] sentMessageSize; |
---|
936 | delete [] recvMessageSize; |
---|
937 | } |
---|
938 | |
---|
939 | Mapper::~Mapper() |
---|
940 | { |
---|
941 | delete [] remapMatrix; |
---|
942 | delete [] srcAddress; |
---|
943 | delete [] srcRank; |
---|
944 | delete [] dstAddress; |
---|
945 | if (neighbourElements) delete [] neighbourElements; |
---|
946 | } |
---|
947 | |
---|
948 | } |
---|