1 |
/* |
2 |
* Copyright (c) 2005 The University of Notre Dame. All Rights Reserved. |
3 |
* |
4 |
* The University of Notre Dame grants you ("Licensee") a |
5 |
* non-exclusive, royalty free, license to use, modify and |
6 |
* redistribute this software in source and binary code form, provided |
7 |
* that the following conditions are met: |
8 |
* |
9 |
* 1. Acknowledgement of the program authors must be made in any |
10 |
* publication of scientific results based in part on use of the |
11 |
* program. An acceptable form of acknowledgement is citation of |
12 |
* the article in which the program was described (Matthew |
13 |
* A. Meineke, Charles F. Vardeman II, Teng Lin, Christopher |
14 |
* J. Fennell and J. Daniel Gezelter, "OOPSE: An Object-Oriented |
15 |
* Parallel Simulation Engine for Molecular Dynamics," |
16 |
* J. Comput. Chem. 26, pp. 252-271 (2005)) |
17 |
* |
18 |
* 2. Redistributions of source code must retain the above copyright |
19 |
* notice, this list of conditions and the following disclaimer. |
20 |
* |
21 |
* 3. Redistributions in binary form must reproduce the above copyright |
22 |
* notice, this list of conditions and the following disclaimer in the |
23 |
* documentation and/or other materials provided with the |
24 |
* distribution. |
25 |
* |
26 |
* This software is provided "AS IS," without a warranty of any |
27 |
* kind. All express or implied conditions, representations and |
28 |
* warranties, including any implied warranty of merchantability, |
29 |
* fitness for a particular purpose or non-infringement, are hereby |
30 |
* excluded. The University of Notre Dame and its licensors shall not |
31 |
* be liable for any damages suffered by licensee as a result of |
32 |
* using, modifying or distributing the software or its |
33 |
* derivatives. In no event will the University of Notre Dame or its |
34 |
* licensors be liable for any lost revenue, profit or data, or for |
35 |
* direct, indirect, special, consequential, incidental or punitive |
36 |
* damages, however caused and regardless of the theory of liability, |
37 |
* arising out of the use of or inability to use software, even if the |
38 |
* University of Notre Dame has been advised of the possibility of |
39 |
* such damages. |
40 |
*/ |
41 |
|
42 |
/** |
43 |
* @file SimCreator.cpp |
44 |
* @author tlin |
45 |
* @date 11/03/2004 |
46 |
* @time 13:51am |
47 |
* @version 1.0 |
48 |
*/ |
49 |
#include <exception> |
50 |
#include <iostream> |
51 |
#include <sstream> |
52 |
#include <string> |
53 |
|
54 |
#include "brains/MoleculeCreator.hpp" |
55 |
#include "brains/SimCreator.hpp" |
56 |
#include "brains/SimSnapshotManager.hpp" |
57 |
#include "io/DumpReader.hpp" |
58 |
#include "UseTheForce/ForceFieldFactory.hpp" |
59 |
#include "utils/simError.h" |
60 |
#include "utils/StringUtils.hpp" |
61 |
#include "math/SeqRandNumGen.hpp" |
62 |
#include "mdParser/MDLexer.hpp" |
63 |
#include "mdParser/MDParser.hpp" |
64 |
#include "mdParser/MDTreeParser.hpp" |
65 |
#include "mdParser/SimplePreprocessor.hpp" |
66 |
#include "antlr/ANTLRException.hpp" |
67 |
#include "antlr/TokenStreamRecognitionException.hpp" |
68 |
#include "antlr/TokenStreamIOException.hpp" |
69 |
#include "antlr/TokenStreamException.hpp" |
70 |
#include "antlr/RecognitionException.hpp" |
71 |
#include "antlr/CharStreamException.hpp" |
72 |
|
73 |
#include "antlr/MismatchedCharException.hpp" |
74 |
#include "antlr/MismatchedTokenException.hpp" |
75 |
#include "antlr/NoViableAltForCharException.hpp" |
76 |
#include "antlr/NoViableAltException.hpp" |
77 |
|
78 |
#ifdef IS_MPI |
79 |
#include "math/ParallelRandNumGen.hpp" |
80 |
#endif |
81 |
|
82 |
namespace oopse { |
83 |
|
84 |
Globals* SimCreator::parseFile(std::istream& rawMetaDataStream, const std::string& filename, int startOfMetaDataBlock ){ |
85 |
Globals* simParams = NULL; |
86 |
try { |
87 |
|
88 |
// Create a preprocessor that preprocesses md file into an ostringstream |
89 |
std::stringstream ppStream; |
90 |
#ifdef IS_MPI |
91 |
int streamSize; |
92 |
const int masterNode = 0; |
93 |
int commStatus; |
94 |
if (worldRank == masterNode) { |
95 |
#endif |
96 |
|
97 |
SimplePreprocessor preprocessor; |
98 |
preprocessor.preprocess(rawMetaDataStream, filename, startOfMetaDataBlock, ppStream); |
99 |
|
100 |
#ifdef IS_MPI |
101 |
//brocasting the stream size |
102 |
streamSize = ppStream.str().size() +1; |
103 |
commStatus = MPI_Bcast(&streamSize, 1, MPI_LONG, masterNode, MPI_COMM_WORLD); |
104 |
|
105 |
commStatus = MPI_Bcast(static_cast<void*>(const_cast<char*>(ppStream.str().c_str())), streamSize, MPI_CHAR, masterNode, MPI_COMM_WORLD); |
106 |
|
107 |
|
108 |
} else { |
109 |
//get stream size |
110 |
commStatus = MPI_Bcast(&streamSize, 1, MPI_LONG, masterNode, MPI_COMM_WORLD); |
111 |
|
112 |
char* buf = new char[streamSize]; |
113 |
assert(buf); |
114 |
|
115 |
//receive file content |
116 |
commStatus = MPI_Bcast(buf, streamSize, MPI_CHAR, masterNode, MPI_COMM_WORLD); |
117 |
|
118 |
ppStream.str(buf); |
119 |
delete buf; |
120 |
|
121 |
} |
122 |
#endif |
123 |
// Create a scanner that reads from the input stream |
124 |
MDLexer lexer(ppStream); |
125 |
lexer.setFilename(filename); |
126 |
lexer.initDeferredLineCount(); |
127 |
|
128 |
// Create a parser that reads from the scanner |
129 |
MDParser parser(lexer); |
130 |
parser.setFilename(filename); |
131 |
|
132 |
// Create an observer that synchorizes file name change |
133 |
FilenameObserver observer; |
134 |
observer.setLexer(&lexer); |
135 |
observer.setParser(&parser); |
136 |
lexer.setObserver(&observer); |
137 |
|
138 |
antlr::ASTFactory factory; |
139 |
parser.initializeASTFactory(factory); |
140 |
parser.setASTFactory(&factory); |
141 |
parser.mdfile(); |
142 |
|
143 |
// Create a tree parser that reads information into Globals |
144 |
MDTreeParser treeParser; |
145 |
treeParser.initializeASTFactory(factory); |
146 |
treeParser.setASTFactory(&factory); |
147 |
simParams = treeParser.walkTree(parser.getAST()); |
148 |
} |
149 |
|
150 |
|
151 |
catch(antlr::MismatchedCharException& e) { |
152 |
sprintf(painCave.errMsg, |
153 |
"parser exception: %s %s:%d:%d\n", |
154 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
155 |
painCave.isFatal = 1; |
156 |
simError(); |
157 |
} |
158 |
catch(antlr::MismatchedTokenException &e) { |
159 |
sprintf(painCave.errMsg, |
160 |
"parser exception: %s %s:%d:%d\n", |
161 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
162 |
painCave.isFatal = 1; |
163 |
simError(); |
164 |
} |
165 |
catch(antlr::NoViableAltForCharException &e) { |
166 |
sprintf(painCave.errMsg, |
167 |
"parser exception: %s %s:%d:%d\n", |
168 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
169 |
painCave.isFatal = 1; |
170 |
simError(); |
171 |
} |
172 |
catch(antlr::NoViableAltException &e) { |
173 |
sprintf(painCave.errMsg, |
174 |
"parser exception: %s %s:%d:%d\n", |
175 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
176 |
painCave.isFatal = 1; |
177 |
simError(); |
178 |
} |
179 |
|
180 |
catch(antlr::TokenStreamRecognitionException& e) { |
181 |
sprintf(painCave.errMsg, |
182 |
"parser exception: %s %s:%d:%d\n", |
183 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
184 |
painCave.isFatal = 1; |
185 |
simError(); |
186 |
} |
187 |
|
188 |
catch(antlr::TokenStreamIOException& e) { |
189 |
sprintf(painCave.errMsg, |
190 |
"parser exception: %s\n", |
191 |
e.getMessage().c_str()); |
192 |
painCave.isFatal = 1; |
193 |
simError(); |
194 |
} |
195 |
|
196 |
catch(antlr::TokenStreamException& e) { |
197 |
sprintf(painCave.errMsg, |
198 |
"parser exception: %s\n", |
199 |
e.getMessage().c_str()); |
200 |
painCave.isFatal = 1; |
201 |
simError(); |
202 |
} |
203 |
catch (antlr::RecognitionException& e) { |
204 |
sprintf(painCave.errMsg, |
205 |
"parser exception: %s %s:%d:%d\n", |
206 |
e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn()); |
207 |
painCave.isFatal = 1; |
208 |
simError(); |
209 |
} |
210 |
catch (antlr::CharStreamException& e) { |
211 |
sprintf(painCave.errMsg, |
212 |
"parser exception: %s\n", |
213 |
e.getMessage().c_str()); |
214 |
painCave.isFatal = 1; |
215 |
simError(); |
216 |
} |
217 |
catch (OOPSEException& e) { |
218 |
sprintf(painCave.errMsg, |
219 |
"%s\n", |
220 |
e.getMessage().c_str()); |
221 |
painCave.isFatal = 1; |
222 |
simError(); |
223 |
} |
224 |
catch (std::exception& e) { |
225 |
sprintf(painCave.errMsg, |
226 |
"parser exception: %s\n", |
227 |
e.what()); |
228 |
painCave.isFatal = 1; |
229 |
simError(); |
230 |
} |
231 |
|
232 |
return simParams; |
233 |
} |
234 |
|
235 |
SimInfo* SimCreator::createSim(const std::string & mdFileName, |
236 |
bool loadInitCoords) { |
237 |
|
238 |
const int bufferSize = 65535; |
239 |
char buffer[bufferSize]; |
240 |
int lineNo = 0; |
241 |
std::string mdRawData; |
242 |
int metaDataBlockStart = -1; |
243 |
int metaDataBlockEnd = -1; |
244 |
int i; |
245 |
int mdOffset; |
246 |
|
247 |
#ifdef IS_MPI |
248 |
const int masterNode = 0; |
249 |
if (worldRank == masterNode) { |
250 |
#endif |
251 |
|
252 |
std::ifstream mdFile_(mdFileName.c_str()); |
253 |
|
254 |
if (mdFile_.fail()) { |
255 |
sprintf(painCave.errMsg, |
256 |
"SimCreator: Cannot open file: %s\n", |
257 |
mdFileName.c_str()); |
258 |
painCave.isFatal = 1; |
259 |
simError(); |
260 |
} |
261 |
|
262 |
mdFile_.getline(buffer, bufferSize); |
263 |
++lineNo; |
264 |
std::string line = trimLeftCopy(buffer); |
265 |
i = CaseInsensitiveFind(line, "<OOPSE"); |
266 |
if (i == string::npos) { |
267 |
sprintf(painCave.errMsg, |
268 |
"SimCreator: File: %s is not an OOPSE file!\n", |
269 |
mdFileName.c_str()); |
270 |
painCave.isFatal = 1; |
271 |
simError(); |
272 |
} |
273 |
|
274 |
//scan through the input stream and find MetaData tag |
275 |
while(mdFile_.getline(buffer, bufferSize)) { |
276 |
++lineNo; |
277 |
|
278 |
std::string line = trimLeftCopy(buffer); |
279 |
if (metaDataBlockStart == -1) { |
280 |
i = CaseInsensitiveFind(line, "<MetaData>"); |
281 |
if (i != string::npos) { |
282 |
metaDataBlockStart = lineNo; |
283 |
mdOffset = mdFile_.tellg(); |
284 |
} |
285 |
} else { |
286 |
i = CaseInsensitiveFind(line, "</MetaData>"); |
287 |
if (i != string::npos) { |
288 |
metaDataBlockEnd = lineNo; |
289 |
} |
290 |
} |
291 |
} |
292 |
|
293 |
if (metaDataBlockStart == -1) { |
294 |
sprintf(painCave.errMsg, |
295 |
"SimCreator: File: %s did not contain a <MetaData> tag!\n", |
296 |
mdFileName.c_str()); |
297 |
painCave.isFatal = 1; |
298 |
simError(); |
299 |
} |
300 |
if (metaDataBlockEnd == -1) { |
301 |
sprintf(painCave.errMsg, |
302 |
"SimCreator: File: %s did not contain a closed MetaData block!\n", |
303 |
mdFileName.c_str()); |
304 |
painCave.isFatal = 1; |
305 |
simError(); |
306 |
} |
307 |
|
308 |
mdFile_.clear(); |
309 |
mdFile_.seekg(0); |
310 |
mdFile_.seekg(mdOffset); |
311 |
|
312 |
mdRawData.clear(); |
313 |
|
314 |
for (int i = 0; i < metaDataBlockEnd - metaDataBlockStart - 1; ++i) { |
315 |
mdFile_.getline(buffer, bufferSize); |
316 |
mdRawData += buffer; |
317 |
mdRawData += "\n"; |
318 |
} |
319 |
|
320 |
mdFile_.close(); |
321 |
|
322 |
#ifdef IS_MPI |
323 |
} |
324 |
#endif |
325 |
|
326 |
std::stringstream rawMetaDataStream(mdRawData); |
327 |
|
328 |
//parse meta-data file |
329 |
Globals* simParams = parseFile(rawMetaDataStream, mdFileName, metaDataBlockStart+1); |
330 |
|
331 |
//create the force field |
332 |
ForceField * ff = ForceFieldFactory::getInstance()->createForceField(simParams->getForceField()); |
333 |
|
334 |
if (ff == NULL) { |
335 |
sprintf(painCave.errMsg, |
336 |
"ForceField Factory can not create %s force field\n", |
337 |
simParams->getForceField().c_str()); |
338 |
painCave.isFatal = 1; |
339 |
simError(); |
340 |
} |
341 |
|
342 |
if (simParams->haveForceFieldFileName()) { |
343 |
ff->setForceFieldFileName(simParams->getForceFieldFileName()); |
344 |
} |
345 |
|
346 |
std::string forcefieldFileName; |
347 |
forcefieldFileName = ff->getForceFieldFileName(); |
348 |
|
349 |
if (simParams->haveForceFieldVariant()) { |
350 |
//If the force field has variant, the variant force field name will be |
351 |
//Base.variant.frc. For exampel EAM.u6.frc |
352 |
|
353 |
std::string variant = simParams->getForceFieldVariant(); |
354 |
|
355 |
std::string::size_type pos = forcefieldFileName.rfind(".frc"); |
356 |
variant = "." + variant; |
357 |
if (pos != std::string::npos) { |
358 |
forcefieldFileName.insert(pos, variant); |
359 |
} else { |
360 |
//If the default force field file name does not containt .frc suffix, just append the .variant |
361 |
forcefieldFileName.append(variant); |
362 |
} |
363 |
} |
364 |
|
365 |
ff->parse(forcefieldFileName); |
366 |
ff->setFortranForceOptions(); |
367 |
//create SimInfo |
368 |
SimInfo * info = new SimInfo(ff, simParams); |
369 |
|
370 |
info->setRawMetaData(mdRawData); |
371 |
|
372 |
//gather parameters (SimCreator only retrieves part of the |
373 |
//parameters) |
374 |
gatherParameters(info, mdFileName); |
375 |
|
376 |
//divide the molecules and determine the global index of molecules |
377 |
#ifdef IS_MPI |
378 |
divideMolecules(info); |
379 |
#endif |
380 |
|
381 |
//create the molecules |
382 |
createMolecules(info); |
383 |
|
384 |
|
385 |
//allocate memory for DataStorage(circular reference, need to |
386 |
//break it) |
387 |
info->setSnapshotManager(new SimSnapshotManager(info)); |
388 |
|
389 |
//set the global index of atoms, rigidbodies and cutoffgroups |
390 |
//(only need to be set once, the global index will never change |
391 |
//again). Local indices of atoms and rigidbodies are already set |
392 |
//by MoleculeCreator class which actually delegates the |
393 |
//responsibility to LocalIndexManager. |
394 |
setGlobalIndex(info); |
395 |
|
396 |
//Although addExcludePairs is called inside SimInfo's addMolecule |
397 |
//method, at that point atoms don't have the global index yet |
398 |
//(their global index are all initialized to -1). Therefore we |
399 |
//have to call addExcludePairs explicitly here. A way to work |
400 |
//around is that we can determine the beginning global indices of |
401 |
//atoms before they get created. |
402 |
SimInfo::MoleculeIterator mi; |
403 |
Molecule* mol; |
404 |
for (mol= info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) { |
405 |
info->addExcludePairs(mol); |
406 |
} |
407 |
|
408 |
if (loadInitCoords) |
409 |
loadCoordinates(info, mdFileName); |
410 |
|
411 |
return info; |
412 |
} |
413 |
|
414 |
void SimCreator::gatherParameters(SimInfo *info, const std::string& mdfile) { |
415 |
|
416 |
//figure out the output file names |
417 |
std::string prefix; |
418 |
|
419 |
#ifdef IS_MPI |
420 |
|
421 |
if (worldRank == 0) { |
422 |
#endif // is_mpi |
423 |
Globals * simParams = info->getSimParams(); |
424 |
if (simParams->haveFinalConfig()) { |
425 |
prefix = getPrefix(simParams->getFinalConfig()); |
426 |
} else { |
427 |
prefix = getPrefix(mdfile); |
428 |
} |
429 |
|
430 |
info->setFinalConfigFileName(prefix + ".eor"); |
431 |
info->setDumpFileName(prefix + ".dump"); |
432 |
info->setStatFileName(prefix + ".stat"); |
433 |
info->setRestFileName(prefix + ".zang"); |
434 |
|
435 |
#ifdef IS_MPI |
436 |
|
437 |
} |
438 |
|
439 |
#endif |
440 |
|
441 |
} |
442 |
|
443 |
#ifdef IS_MPI |
444 |
void SimCreator::divideMolecules(SimInfo *info) { |
445 |
RealType numerator; |
446 |
RealType denominator; |
447 |
RealType precast; |
448 |
RealType x; |
449 |
RealType y; |
450 |
RealType a; |
451 |
int old_atoms; |
452 |
int add_atoms; |
453 |
int new_atoms; |
454 |
int nTarget; |
455 |
int done; |
456 |
int i; |
457 |
int j; |
458 |
int loops; |
459 |
int which_proc; |
460 |
int nProcessors; |
461 |
std::vector<int> atomsPerProc; |
462 |
int nGlobalMols = info->getNGlobalMolecules(); |
463 |
std::vector<int> molToProcMap(nGlobalMols, -1); // default to an error condition: |
464 |
|
465 |
MPI_Comm_size(MPI_COMM_WORLD, &nProcessors); |
466 |
|
467 |
if (nProcessors > nGlobalMols) { |
468 |
sprintf(painCave.errMsg, |
469 |
"nProcessors (%d) > nMol (%d)\n" |
470 |
"\tThe number of processors is larger than\n" |
471 |
"\tthe number of molecules. This will not result in a \n" |
472 |
"\tusable division of atoms for force decomposition.\n" |
473 |
"\tEither try a smaller number of processors, or run the\n" |
474 |
"\tsingle-processor version of OOPSE.\n", nProcessors, nGlobalMols); |
475 |
|
476 |
painCave.isFatal = 1; |
477 |
simError(); |
478 |
} |
479 |
|
480 |
int seedValue; |
481 |
Globals * simParams = info->getSimParams(); |
482 |
SeqRandNumGen* myRandom; //divide labor does not need Parallel random number generator |
483 |
if (simParams->haveSeed()) { |
484 |
seedValue = simParams->getSeed(); |
485 |
myRandom = new SeqRandNumGen(seedValue); |
486 |
}else { |
487 |
myRandom = new SeqRandNumGen(); |
488 |
} |
489 |
|
490 |
|
491 |
a = 3.0 * nGlobalMols / info->getNGlobalAtoms(); |
492 |
|
493 |
//initialize atomsPerProc |
494 |
atomsPerProc.insert(atomsPerProc.end(), nProcessors, 0); |
495 |
|
496 |
if (worldRank == 0) { |
497 |
numerator = info->getNGlobalAtoms(); |
498 |
denominator = nProcessors; |
499 |
precast = numerator / denominator; |
500 |
nTarget = (int)(precast + 0.5); |
501 |
|
502 |
for(i = 0; i < nGlobalMols; i++) { |
503 |
done = 0; |
504 |
loops = 0; |
505 |
|
506 |
while (!done) { |
507 |
loops++; |
508 |
|
509 |
// Pick a processor at random |
510 |
|
511 |
which_proc = (int) (myRandom->rand() * nProcessors); |
512 |
|
513 |
//get the molecule stamp first |
514 |
int stampId = info->getMoleculeStampId(i); |
515 |
MoleculeStamp * moleculeStamp = info->getMoleculeStamp(stampId); |
516 |
|
517 |
// How many atoms does this processor have so far? |
518 |
old_atoms = atomsPerProc[which_proc]; |
519 |
add_atoms = moleculeStamp->getNAtoms(); |
520 |
new_atoms = old_atoms + add_atoms; |
521 |
|
522 |
// If we've been through this loop too many times, we need |
523 |
// to just give up and assign the molecule to this processor |
524 |
// and be done with it. |
525 |
|
526 |
if (loops > 100) { |
527 |
sprintf(painCave.errMsg, |
528 |
"I've tried 100 times to assign molecule %d to a " |
529 |
" processor, but can't find a good spot.\n" |
530 |
"I'm assigning it at random to processor %d.\n", |
531 |
i, which_proc); |
532 |
|
533 |
painCave.isFatal = 0; |
534 |
simError(); |
535 |
|
536 |
molToProcMap[i] = which_proc; |
537 |
atomsPerProc[which_proc] += add_atoms; |
538 |
|
539 |
done = 1; |
540 |
continue; |
541 |
} |
542 |
|
543 |
// If we can add this molecule to this processor without sending |
544 |
// it above nTarget, then go ahead and do it: |
545 |
|
546 |
if (new_atoms <= nTarget) { |
547 |
molToProcMap[i] = which_proc; |
548 |
atomsPerProc[which_proc] += add_atoms; |
549 |
|
550 |
done = 1; |
551 |
continue; |
552 |
} |
553 |
|
554 |
// The only situation left is when new_atoms > nTarget. We |
555 |
// want to accept this with some probability that dies off the |
556 |
// farther we are from nTarget |
557 |
|
558 |
// roughly: x = new_atoms - nTarget |
559 |
// Pacc(x) = exp(- a * x) |
560 |
// where a = penalty / (average atoms per molecule) |
561 |
|
562 |
x = (RealType)(new_atoms - nTarget); |
563 |
y = myRandom->rand(); |
564 |
|
565 |
if (y < exp(- a * x)) { |
566 |
molToProcMap[i] = which_proc; |
567 |
atomsPerProc[which_proc] += add_atoms; |
568 |
|
569 |
done = 1; |
570 |
continue; |
571 |
} else { |
572 |
continue; |
573 |
} |
574 |
} |
575 |
} |
576 |
|
577 |
delete myRandom; |
578 |
|
579 |
// Spray out this nonsense to all other processors: |
580 |
|
581 |
MPI_Bcast(&molToProcMap[0], nGlobalMols, MPI_INT, 0, MPI_COMM_WORLD); |
582 |
} else { |
583 |
|
584 |
// Listen to your marching orders from processor 0: |
585 |
|
586 |
MPI_Bcast(&molToProcMap[0], nGlobalMols, MPI_INT, 0, MPI_COMM_WORLD); |
587 |
} |
588 |
|
589 |
info->setMolToProcMap(molToProcMap); |
590 |
sprintf(checkPointMsg, |
591 |
"Successfully divided the molecules among the processors.\n"); |
592 |
errorCheckPoint(); |
593 |
} |
594 |
|
595 |
#endif |
596 |
|
597 |
void SimCreator::createMolecules(SimInfo *info) { |
598 |
MoleculeCreator molCreator; |
599 |
int stampId; |
600 |
|
601 |
for(int i = 0; i < info->getNGlobalMolecules(); i++) { |
602 |
|
603 |
#ifdef IS_MPI |
604 |
|
605 |
if (info->getMolToProc(i) == worldRank) { |
606 |
#endif |
607 |
|
608 |
stampId = info->getMoleculeStampId(i); |
609 |
Molecule * mol = molCreator.createMolecule(info->getForceField(), info->getMoleculeStamp(stampId), |
610 |
stampId, i, info->getLocalIndexManager()); |
611 |
|
612 |
info->addMolecule(mol); |
613 |
|
614 |
#ifdef IS_MPI |
615 |
|
616 |
} |
617 |
|
618 |
#endif |
619 |
|
620 |
} //end for(int i=0) |
621 |
} |
622 |
|
623 |
void SimCreator::setGlobalIndex(SimInfo *info) { |
624 |
SimInfo::MoleculeIterator mi; |
625 |
Molecule::AtomIterator ai; |
626 |
Molecule::RigidBodyIterator ri; |
627 |
Molecule::CutoffGroupIterator ci; |
628 |
Molecule::IntegrableObjectIterator ioi; |
629 |
Molecule * mol; |
630 |
Atom * atom; |
631 |
RigidBody * rb; |
632 |
CutoffGroup * cg; |
633 |
int beginAtomIndex; |
634 |
int beginRigidBodyIndex; |
635 |
int beginCutoffGroupIndex; |
636 |
int nGlobalAtoms = info->getNGlobalAtoms(); |
637 |
|
638 |
/**@todo fixme */ |
639 |
#ifndef IS_MPI |
640 |
|
641 |
beginAtomIndex = 0; |
642 |
beginRigidBodyIndex = 0; |
643 |
beginCutoffGroupIndex = 0; |
644 |
|
645 |
#else |
646 |
|
647 |
int nproc; |
648 |
int myNode; |
649 |
|
650 |
myNode = worldRank; |
651 |
MPI_Comm_size(MPI_COMM_WORLD, &nproc); |
652 |
|
653 |
std::vector < int > tmpAtomsInProc(nproc, 0); |
654 |
std::vector < int > tmpRigidBodiesInProc(nproc, 0); |
655 |
std::vector < int > tmpCutoffGroupsInProc(nproc, 0); |
656 |
std::vector < int > NumAtomsInProc(nproc, 0); |
657 |
std::vector < int > NumRigidBodiesInProc(nproc, 0); |
658 |
std::vector < int > NumCutoffGroupsInProc(nproc, 0); |
659 |
|
660 |
tmpAtomsInProc[myNode] = info->getNAtoms(); |
661 |
tmpRigidBodiesInProc[myNode] = info->getNRigidBodies(); |
662 |
tmpCutoffGroupsInProc[myNode] = info->getNCutoffGroups(); |
663 |
|
664 |
//do MPI_ALLREDUCE to exchange the total number of atoms, rigidbodies and cutoff groups |
665 |
MPI_Allreduce(&tmpAtomsInProc[0], &NumAtomsInProc[0], nproc, MPI_INT, |
666 |
MPI_SUM, MPI_COMM_WORLD); |
667 |
MPI_Allreduce(&tmpRigidBodiesInProc[0], &NumRigidBodiesInProc[0], nproc, |
668 |
MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
669 |
MPI_Allreduce(&tmpCutoffGroupsInProc[0], &NumCutoffGroupsInProc[0], nproc, |
670 |
MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
671 |
|
672 |
beginAtomIndex = 0; |
673 |
beginRigidBodyIndex = 0; |
674 |
beginCutoffGroupIndex = 0; |
675 |
|
676 |
for(int i = 0; i < myNode; i++) { |
677 |
beginAtomIndex += NumAtomsInProc[i]; |
678 |
beginRigidBodyIndex += NumRigidBodiesInProc[i]; |
679 |
beginCutoffGroupIndex += NumCutoffGroupsInProc[i]; |
680 |
} |
681 |
|
682 |
#endif |
683 |
|
684 |
//rigidbody's index begins right after atom's |
685 |
beginRigidBodyIndex += info->getNGlobalAtoms(); |
686 |
|
687 |
for(mol = info->beginMolecule(mi); mol != NULL; |
688 |
mol = info->nextMolecule(mi)) { |
689 |
|
690 |
//local index(index in DataStorge) of atom is important |
691 |
for(atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) { |
692 |
atom->setGlobalIndex(beginAtomIndex++); |
693 |
} |
694 |
|
695 |
for(rb = mol->beginRigidBody(ri); rb != NULL; |
696 |
rb = mol->nextRigidBody(ri)) { |
697 |
rb->setGlobalIndex(beginRigidBodyIndex++); |
698 |
} |
699 |
|
700 |
//local index of cutoff group is trivial, it only depends on the order of travesing |
701 |
for(cg = mol->beginCutoffGroup(ci); cg != NULL; |
702 |
cg = mol->nextCutoffGroup(ci)) { |
703 |
cg->setGlobalIndex(beginCutoffGroupIndex++); |
704 |
} |
705 |
} |
706 |
|
707 |
//fill globalGroupMembership |
708 |
std::vector<int> globalGroupMembership(info->getNGlobalAtoms(), 0); |
709 |
for(mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) { |
710 |
for (cg = mol->beginCutoffGroup(ci); cg != NULL; cg = mol->nextCutoffGroup(ci)) { |
711 |
|
712 |
for(atom = cg->beginAtom(ai); atom != NULL; atom = cg->nextAtom(ai)) { |
713 |
globalGroupMembership[atom->getGlobalIndex()] = cg->getGlobalIndex(); |
714 |
} |
715 |
|
716 |
} |
717 |
} |
718 |
|
719 |
#ifdef IS_MPI |
720 |
// Since the globalGroupMembership has been zero filled and we've only |
721 |
// poked values into the atoms we know, we can do an Allreduce |
722 |
// to get the full globalGroupMembership array (We think). |
723 |
// This would be prettier if we could use MPI_IN_PLACE like the MPI-2 |
724 |
// docs said we could. |
725 |
std::vector<int> tmpGroupMembership(nGlobalAtoms, 0); |
726 |
MPI_Allreduce(&globalGroupMembership[0], &tmpGroupMembership[0], nGlobalAtoms, |
727 |
MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
728 |
info->setGlobalGroupMembership(tmpGroupMembership); |
729 |
#else |
730 |
info->setGlobalGroupMembership(globalGroupMembership); |
731 |
#endif |
732 |
|
733 |
//fill molMembership |
734 |
std::vector<int> globalMolMembership(info->getNGlobalAtoms(), 0); |
735 |
|
736 |
for(mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) { |
737 |
|
738 |
for(atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) { |
739 |
globalMolMembership[atom->getGlobalIndex()] = mol->getGlobalIndex(); |
740 |
} |
741 |
} |
742 |
|
743 |
#ifdef IS_MPI |
744 |
std::vector<int> tmpMolMembership(nGlobalAtoms, 0); |
745 |
|
746 |
MPI_Allreduce(&globalMolMembership[0], &tmpMolMembership[0], nGlobalAtoms, |
747 |
MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
748 |
|
749 |
info->setGlobalMolMembership(tmpMolMembership); |
750 |
#else |
751 |
info->setGlobalMolMembership(globalMolMembership); |
752 |
#endif |
753 |
|
754 |
// nIOPerMol holds the number of integrable objects per molecule |
755 |
// here the molecules are listed by their global indices. |
756 |
|
757 |
std::vector<int> nIOPerMol(info->getNGlobalMolecules(), 0); |
758 |
for (mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) { |
759 |
nIOPerMol[mol->getGlobalIndex()] = mol->getNIntegrableObjects(); |
760 |
} |
761 |
|
762 |
#ifdef IS_MPI |
763 |
std::vector<int> numIntegrableObjectsPerMol(info->getNGlobalMolecules(), 0); |
764 |
MPI_Allreduce(&nIOPerMol[0], &numIntegrableObjectsPerMol[0], |
765 |
info->getNGlobalMolecules(), MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
766 |
#else |
767 |
std::vector<int> numIntegrableObjectsPerMol = nIOPerMol; |
768 |
#endif |
769 |
|
770 |
std::vector<int> startingIOIndexForMol(info->getNGlobalMolecules()); |
771 |
|
772 |
int startingIndex = 0; |
773 |
for (int i = 0; i < info->getNGlobalMolecules(); i++) { |
774 |
startingIOIndexForMol[i] = startingIndex; |
775 |
startingIndex += numIntegrableObjectsPerMol[i]; |
776 |
} |
777 |
|
778 |
std::vector<StuntDouble*> IOIndexToIntegrableObject(info->getNGlobalIntegrableObjects(), (StuntDouble*)NULL); |
779 |
for (mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) { |
780 |
int myGlobalIndex = mol->getGlobalIndex(); |
781 |
int globalIO = startingIOIndexForMol[myGlobalIndex]; |
782 |
for (StuntDouble* integrableObject = mol->beginIntegrableObject(ioi); integrableObject != NULL; |
783 |
integrableObject = mol->nextIntegrableObject(ioi)) { |
784 |
integrableObject->setGlobalIntegrableObjectIndex(globalIO); |
785 |
IOIndexToIntegrableObject[globalIO] = integrableObject; |
786 |
globalIO++; |
787 |
} |
788 |
} |
789 |
|
790 |
info->setIOIndexToIntegrableObject(IOIndexToIntegrableObject); |
791 |
|
792 |
} |
793 |
|
794 |
void SimCreator::loadCoordinates(SimInfo* info, const std::string& mdFileName) { |
795 |
Globals* simParams; |
796 |
simParams = info->getSimParams(); |
797 |
|
798 |
|
799 |
DumpReader reader(info, mdFileName); |
800 |
int nframes = reader.getNFrames(); |
801 |
|
802 |
if (nframes > 0) { |
803 |
reader.readFrame(nframes - 1); |
804 |
} else { |
805 |
//invalid initial coordinate file |
806 |
sprintf(painCave.errMsg, |
807 |
"Initial configuration file %s should at least contain one frame\n", |
808 |
mdFileName.c_str()); |
809 |
painCave.isFatal = 1; |
810 |
simError(); |
811 |
} |
812 |
|
813 |
//copy the current snapshot to previous snapshot |
814 |
info->getSnapshotManager()->advance(); |
815 |
} |
816 |
|
817 |
} //end namespace oopse |
818 |
|
819 |
|