1 |
< |
#include <stdlib.h> |
2 |
< |
#include <string.h> |
3 |
< |
#include <math.h> |
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. Redistributions of source code must retain the above copyright |
10 |
> |
* notice, this list of conditions and the following disclaimer. |
11 |
> |
* |
12 |
> |
* 2. Redistributions in binary form must reproduce the above copyright |
13 |
> |
* notice, this list of conditions and the following disclaimer in the |
14 |
> |
* documentation and/or other materials provided with the |
15 |
> |
* distribution. |
16 |
> |
* |
17 |
> |
* This software is provided "AS IS," without a warranty of any |
18 |
> |
* kind. All express or implied conditions, representations and |
19 |
> |
* warranties, including any implied warranty of merchantability, |
20 |
> |
* fitness for a particular purpose or non-infringement, are hereby |
21 |
> |
* excluded. The University of Notre Dame and its licensors shall not |
22 |
> |
* be liable for any damages suffered by licensee as a result of |
23 |
> |
* using, modifying or distributing the software or its |
24 |
> |
* derivatives. In no event will the University of Notre Dame or its |
25 |
> |
* licensors be liable for any lost revenue, profit or data, or for |
26 |
> |
* direct, indirect, special, consequential, incidental or punitive |
27 |
> |
* damages, however caused and regardless of the theory of liability, |
28 |
> |
* arising out of the use of or inability to use software, even if the |
29 |
> |
* University of Notre Dame has been advised of the possibility of |
30 |
> |
* such damages. |
31 |
> |
* |
32 |
> |
* SUPPORT OPEN SCIENCE! If you use OpenMD or its source code in your |
33 |
> |
* research, please cite the appropriate papers when you publish your |
34 |
> |
* work. Good starting points are: |
35 |
> |
* |
36 |
> |
* [1] Meineke, et al., J. Comp. Chem. 26, 252-271 (2005). |
37 |
> |
* [2] Fennell & Gezelter, J. Chem. Phys. 124, 234104 (2006). |
38 |
> |
* [3] Sun, Lin & Gezelter, J. Chem. Phys. 128, 24107 (2008). |
39 |
> |
* [4] Kuang & Gezelter, J. Chem. Phys. 133, 164101 (2010). |
40 |
> |
* [5] Vardeman, Stocker & Gezelter, J. Chem. Theory Comput. 7, 834 (2011). |
41 |
> |
*/ |
42 |
> |
|
43 |
> |
/** |
44 |
> |
* @file SimInfo.cpp |
45 |
> |
* @author tlin |
46 |
> |
* @date 11/02/2004 |
47 |
> |
* @version 1.0 |
48 |
> |
*/ |
49 |
|
|
50 |
< |
#include <iostream> |
51 |
< |
using namespace std; |
50 |
> |
#include <algorithm> |
51 |
> |
#include <set> |
52 |
> |
#include <map> |
53 |
|
|
54 |
|
#include "brains/SimInfo.hpp" |
55 |
< |
#define __C |
56 |
< |
#include "brains/fSimulation.h" |
55 |
> |
#include "math/Vector3.hpp" |
56 |
> |
#include "primitives/Molecule.hpp" |
57 |
> |
#include "primitives/StuntDouble.hpp" |
58 |
> |
#include "utils/MemoryUtils.hpp" |
59 |
|
#include "utils/simError.h" |
60 |
< |
#include "UseTheForce/DarkSide/simulation_interface.h" |
61 |
< |
#include "UseTheForce/notifyCutoffs_interface.h" |
62 |
< |
|
63 |
< |
//#include "UseTheForce/fortranWrappers.hpp" |
16 |
< |
|
17 |
< |
#include "math/MatVec3.h" |
18 |
< |
|
60 |
> |
#include "selection/SelectionManager.hpp" |
61 |
> |
#include "io/ForceFieldOptions.hpp" |
62 |
> |
#include "brains/ForceField.hpp" |
63 |
> |
#include "nonbonded/SwitchingFunction.hpp" |
64 |
|
#ifdef IS_MPI |
65 |
< |
#include "brains/mpiSimulation.hpp" |
65 |
> |
#include <mpi.h> |
66 |
|
#endif |
67 |
|
|
68 |
< |
inline double roundMe( double x ){ |
69 |
< |
return ( x >= 0 ) ? floor( x + 0.5 ) : ceil( x - 0.5 ); |
70 |
< |
} |
71 |
< |
|
72 |
< |
inline double min( double a, double b ){ |
73 |
< |
return (a < b ) ? a : b; |
74 |
< |
} |
68 |
> |
using namespace std; |
69 |
> |
namespace OpenMD { |
70 |
> |
|
71 |
> |
SimInfo::SimInfo(ForceField* ff, Globals* simParams) : |
72 |
> |
forceField_(ff), simParams_(simParams), |
73 |
> |
ndf_(0), fdf_local(0), ndfRaw_(0), ndfTrans_(0), nZconstraint_(0), |
74 |
> |
nGlobalMols_(0), nGlobalAtoms_(0), nGlobalCutoffGroups_(0), |
75 |
> |
nGlobalIntegrableObjects_(0), nGlobalRigidBodies_(0), nGlobalFluctuatingCharges_(0), |
76 |
> |
nAtoms_(0), nBonds_(0), nBends_(0), nTorsions_(0), nInversions_(0), |
77 |
> |
nRigidBodies_(0), nIntegrableObjects_(0), nCutoffGroups_(0), |
78 |
> |
nConstraints_(0), nFluctuatingCharges_(0), sman_(NULL), topologyDone_(false), |
79 |
> |
calcBoxDipole_(false), useAtomicVirial_(true) { |
80 |
> |
|
81 |
> |
MoleculeStamp* molStamp; |
82 |
> |
int nMolWithSameStamp; |
83 |
> |
int nCutoffAtoms = 0; // number of atoms belong to cutoff groups |
84 |
> |
int nGroups = 0; //total cutoff groups defined in meta-data file |
85 |
> |
CutoffGroupStamp* cgStamp; |
86 |
> |
RigidBodyStamp* rbStamp; |
87 |
> |
int nRigidAtoms = 0; |
88 |
> |
|
89 |
> |
vector<Component*> components = simParams->getComponents(); |
90 |
> |
|
91 |
> |
for (vector<Component*>::iterator i = components.begin(); |
92 |
> |
i !=components.end(); ++i) { |
93 |
> |
molStamp = (*i)->getMoleculeStamp(); |
94 |
> |
nMolWithSameStamp = (*i)->getNMol(); |
95 |
> |
|
96 |
> |
addMoleculeStamp(molStamp, nMolWithSameStamp); |
97 |
> |
|
98 |
> |
//calculate atoms in molecules |
99 |
> |
nGlobalAtoms_ += molStamp->getNAtoms() *nMolWithSameStamp; |
100 |
> |
|
101 |
> |
//calculate atoms in cutoff groups |
102 |
> |
int nAtomsInGroups = 0; |
103 |
> |
int nCutoffGroupsInStamp = molStamp->getNCutoffGroups(); |
104 |
> |
|
105 |
> |
for (int j=0; j < nCutoffGroupsInStamp; j++) { |
106 |
> |
cgStamp = molStamp->getCutoffGroupStamp(j); |
107 |
> |
nAtomsInGroups += cgStamp->getNMembers(); |
108 |
> |
} |
109 |
> |
|
110 |
> |
nGroups += nCutoffGroupsInStamp * nMolWithSameStamp; |
111 |
> |
|
112 |
> |
nCutoffAtoms += nAtomsInGroups * nMolWithSameStamp; |
113 |
> |
|
114 |
> |
//calculate atoms in rigid bodies |
115 |
> |
int nAtomsInRigidBodies = 0; |
116 |
> |
int nRigidBodiesInStamp = molStamp->getNRigidBodies(); |
117 |
> |
|
118 |
> |
for (int j=0; j < nRigidBodiesInStamp; j++) { |
119 |
> |
rbStamp = molStamp->getRigidBodyStamp(j); |
120 |
> |
nAtomsInRigidBodies += rbStamp->getNMembers(); |
121 |
> |
} |
122 |
> |
|
123 |
> |
nGlobalRigidBodies_ += nRigidBodiesInStamp * nMolWithSameStamp; |
124 |
> |
nRigidAtoms += nAtomsInRigidBodies * nMolWithSameStamp; |
125 |
> |
|
126 |
> |
} |
127 |
> |
|
128 |
> |
//every free atom (atom does not belong to cutoff groups) is a cutoff |
129 |
> |
//group therefore the total number of cutoff groups in the system is |
130 |
> |
//equal to the total number of atoms minus number of atoms belong to |
131 |
> |
//cutoff group defined in meta-data file plus the number of cutoff |
132 |
> |
//groups defined in meta-data file |
133 |
|
|
134 |
< |
SimInfo* currentInfo; |
134 |
> |
nGlobalCutoffGroups_ = nGlobalAtoms_ - nCutoffAtoms + nGroups; |
135 |
> |
|
136 |
> |
//every free atom (atom does not belong to rigid bodies) is an |
137 |
> |
//integrable object therefore the total number of integrable objects |
138 |
> |
//in the system is equal to the total number of atoms minus number of |
139 |
> |
//atoms belong to rigid body defined in meta-data file plus the number |
140 |
> |
//of rigid bodies defined in meta-data file |
141 |
> |
nGlobalIntegrableObjects_ = nGlobalAtoms_ - nRigidAtoms |
142 |
> |
+ nGlobalRigidBodies_; |
143 |
> |
|
144 |
> |
nGlobalMols_ = molStampIds_.size(); |
145 |
> |
molToProcMap_.resize(nGlobalMols_); |
146 |
> |
} |
147 |
> |
|
148 |
> |
SimInfo::~SimInfo() { |
149 |
> |
map<int, Molecule*>::iterator i; |
150 |
> |
for (i = molecules_.begin(); i != molecules_.end(); ++i) { |
151 |
> |
delete i->second; |
152 |
> |
} |
153 |
> |
molecules_.clear(); |
154 |
> |
|
155 |
> |
delete sman_; |
156 |
> |
delete simParams_; |
157 |
> |
delete forceField_; |
158 |
> |
} |
159 |
|
|
33 |
– |
SimInfo::SimInfo(){ |
160 |
|
|
161 |
< |
n_constraints = 0; |
162 |
< |
nZconstraints = 0; |
163 |
< |
n_oriented = 0; |
164 |
< |
n_dipoles = 0; |
165 |
< |
ndf = 0; |
166 |
< |
ndfRaw = 0; |
167 |
< |
nZconstraints = 0; |
168 |
< |
the_integrator = NULL; |
169 |
< |
setTemp = 0; |
170 |
< |
thermalTime = 0.0; |
171 |
< |
currentTime = 0.0; |
172 |
< |
rCut = 0.0; |
173 |
< |
rSw = 0.0; |
174 |
< |
|
175 |
< |
haveRcut = 0; |
176 |
< |
haveRsw = 0; |
177 |
< |
boxIsInit = 0; |
161 |
> |
bool SimInfo::addMolecule(Molecule* mol) { |
162 |
> |
MoleculeIterator i; |
163 |
> |
|
164 |
> |
i = molecules_.find(mol->getGlobalIndex()); |
165 |
> |
if (i == molecules_.end() ) { |
166 |
> |
|
167 |
> |
molecules_.insert(make_pair(mol->getGlobalIndex(), mol)); |
168 |
> |
|
169 |
> |
nAtoms_ += mol->getNAtoms(); |
170 |
> |
nBonds_ += mol->getNBonds(); |
171 |
> |
nBends_ += mol->getNBends(); |
172 |
> |
nTorsions_ += mol->getNTorsions(); |
173 |
> |
nInversions_ += mol->getNInversions(); |
174 |
> |
nRigidBodies_ += mol->getNRigidBodies(); |
175 |
> |
nIntegrableObjects_ += mol->getNIntegrableObjects(); |
176 |
> |
nCutoffGroups_ += mol->getNCutoffGroups(); |
177 |
> |
nConstraints_ += mol->getNConstraintPairs(); |
178 |
> |
|
179 |
> |
addInteractionPairs(mol); |
180 |
> |
|
181 |
> |
return true; |
182 |
> |
} else { |
183 |
> |
return false; |
184 |
> |
} |
185 |
> |
} |
186 |
|
|
187 |
< |
resetTime = 1e99; |
187 |
> |
bool SimInfo::removeMolecule(Molecule* mol) { |
188 |
> |
MoleculeIterator i; |
189 |
> |
i = molecules_.find(mol->getGlobalIndex()); |
190 |
|
|
191 |
< |
orthoRhombic = 0; |
56 |
< |
orthoTolerance = 1E-6; |
57 |
< |
useInitXSstate = true; |
191 |
> |
if (i != molecules_.end() ) { |
192 |
|
|
193 |
< |
usePBC = 0; |
194 |
< |
useDirectionalAtoms = 0; |
195 |
< |
useLennardJones = 0; |
196 |
< |
useElectrostatics = 0; |
197 |
< |
useCharges = 0; |
198 |
< |
useDipoles = 0; |
199 |
< |
useSticky = 0; |
200 |
< |
useGayBerne = 0; |
201 |
< |
useEAM = 0; |
202 |
< |
useShapes = 0; |
203 |
< |
useFLARB = 0; |
193 |
> |
assert(mol == i->second); |
194 |
> |
|
195 |
> |
nAtoms_ -= mol->getNAtoms(); |
196 |
> |
nBonds_ -= mol->getNBonds(); |
197 |
> |
nBends_ -= mol->getNBends(); |
198 |
> |
nTorsions_ -= mol->getNTorsions(); |
199 |
> |
nInversions_ -= mol->getNInversions(); |
200 |
> |
nRigidBodies_ -= mol->getNRigidBodies(); |
201 |
> |
nIntegrableObjects_ -= mol->getNIntegrableObjects(); |
202 |
> |
nCutoffGroups_ -= mol->getNCutoffGroups(); |
203 |
> |
nConstraints_ -= mol->getNConstraintPairs(); |
204 |
|
|
205 |
< |
useSolidThermInt = 0; |
206 |
< |
useLiquidThermInt = 0; |
205 |
> |
removeInteractionPairs(mol); |
206 |
> |
molecules_.erase(mol->getGlobalIndex()); |
207 |
|
|
208 |
< |
haveCutoffGroups = false; |
208 |
> |
delete mol; |
209 |
> |
|
210 |
> |
return true; |
211 |
> |
} else { |
212 |
> |
return false; |
213 |
> |
} |
214 |
> |
} |
215 |
|
|
216 |
< |
excludes = Exclude::Instance(); |
216 |
> |
|
217 |
> |
Molecule* SimInfo::beginMolecule(MoleculeIterator& i) { |
218 |
> |
i = molecules_.begin(); |
219 |
> |
return i == molecules_.end() ? NULL : i->second; |
220 |
> |
} |
221 |
|
|
222 |
< |
myConfiguration = new SimState(); |
222 |
> |
Molecule* SimInfo::nextMolecule(MoleculeIterator& i) { |
223 |
> |
++i; |
224 |
> |
return i == molecules_.end() ? NULL : i->second; |
225 |
> |
} |
226 |
|
|
80 |
– |
has_minimizer = false; |
81 |
– |
the_minimizer =NULL; |
227 |
|
|
228 |
< |
ngroup = 0; |
228 |
> |
void SimInfo::calcNdf() { |
229 |
> |
int ndf_local, nfq_local; |
230 |
> |
MoleculeIterator i; |
231 |
> |
vector<StuntDouble*>::iterator j; |
232 |
> |
vector<Atom*>::iterator k; |
233 |
|
|
234 |
< |
} |
234 |
> |
Molecule* mol; |
235 |
> |
StuntDouble* sd; |
236 |
> |
Atom* atom; |
237 |
|
|
238 |
+ |
ndf_local = 0; |
239 |
+ |
nfq_local = 0; |
240 |
+ |
|
241 |
+ |
for (mol = beginMolecule(i); mol != NULL; mol = nextMolecule(i)) { |
242 |
|
|
243 |
< |
SimInfo::~SimInfo(){ |
243 |
> |
for (sd = mol->beginIntegrableObject(j); sd != NULL; |
244 |
> |
sd = mol->nextIntegrableObject(j)) { |
245 |
|
|
246 |
< |
delete myConfiguration; |
246 |
> |
ndf_local += 3; |
247 |
|
|
248 |
< |
map<string, GenericData*>::iterator i; |
249 |
< |
|
250 |
< |
for(i = properties.begin(); i != properties.end(); i++) |
251 |
< |
delete (*i).second; |
248 |
> |
if (sd->isDirectional()) { |
249 |
> |
if (sd->isLinear()) { |
250 |
> |
ndf_local += 2; |
251 |
> |
} else { |
252 |
> |
ndf_local += 3; |
253 |
> |
} |
254 |
> |
} |
255 |
> |
} |
256 |
|
|
257 |
< |
} |
257 |
> |
for (atom = mol->beginFluctuatingCharge(k); atom != NULL; |
258 |
> |
atom = mol->nextFluctuatingCharge(k)) { |
259 |
> |
if (atom->isFluctuatingCharge()) { |
260 |
> |
nfq_local++; |
261 |
> |
} |
262 |
> |
} |
263 |
> |
} |
264 |
> |
|
265 |
> |
ndfLocal_ = ndf_local; |
266 |
|
|
267 |
< |
void SimInfo::setBox(double newBox[3]) { |
268 |
< |
|
101 |
< |
int i, j; |
102 |
< |
double tempMat[3][3]; |
267 |
> |
// n_constraints is local, so subtract them on each processor |
268 |
> |
ndf_local -= nConstraints_; |
269 |
|
|
270 |
< |
for(i=0; i<3; i++) |
271 |
< |
for (j=0; j<3; j++) tempMat[i][j] = 0.0;; |
270 |
> |
#ifdef IS_MPI |
271 |
> |
MPI_Allreduce(&ndf_local,&ndf_,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
272 |
> |
MPI_Allreduce(&nfq_local,&nGlobalFluctuatingCharges_,1, MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
273 |
> |
#else |
274 |
> |
ndf_ = ndf_local; |
275 |
> |
nGlobalFluctuatingCharges_ = nfq_local; |
276 |
> |
#endif |
277 |
|
|
278 |
< |
tempMat[0][0] = newBox[0]; |
279 |
< |
tempMat[1][1] = newBox[1]; |
280 |
< |
tempMat[2][2] = newBox[2]; |
278 |
> |
// nZconstraints_ is global, as are the 3 COM translations for the |
279 |
> |
// entire system: |
280 |
> |
ndf_ = ndf_ - 3 - nZconstraint_; |
281 |
|
|
282 |
< |
setBoxM( tempMat ); |
282 |
> |
} |
283 |
|
|
284 |
< |
} |
285 |
< |
|
286 |
< |
void SimInfo::setBoxM( double theBox[3][3] ){ |
284 |
> |
int SimInfo::getFdf() { |
285 |
> |
#ifdef IS_MPI |
286 |
> |
MPI_Allreduce(&fdf_local,&fdf_,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
287 |
> |
#else |
288 |
> |
fdf_ = fdf_local; |
289 |
> |
#endif |
290 |
> |
return fdf_; |
291 |
> |
} |
292 |
|
|
293 |
< |
int i, j; |
294 |
< |
double FortranHmat[9]; // to preserve compatibility with Fortran the |
295 |
< |
// ordering in the array is as follows: |
296 |
< |
// [ 0 3 6 ] |
297 |
< |
// [ 1 4 7 ] |
298 |
< |
// [ 2 5 8 ] |
299 |
< |
double FortranHmatInv[9]; // the inverted Hmat (for Fortran); |
293 |
> |
unsigned int SimInfo::getNLocalCutoffGroups(){ |
294 |
> |
int nLocalCutoffAtoms = 0; |
295 |
> |
Molecule* mol; |
296 |
> |
MoleculeIterator mi; |
297 |
> |
CutoffGroup* cg; |
298 |
> |
Molecule::CutoffGroupIterator ci; |
299 |
> |
|
300 |
> |
for (mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
301 |
> |
|
302 |
> |
for (cg = mol->beginCutoffGroup(ci); cg != NULL; |
303 |
> |
cg = mol->nextCutoffGroup(ci)) { |
304 |
> |
nLocalCutoffAtoms += cg->getNumAtom(); |
305 |
> |
|
306 |
> |
} |
307 |
> |
} |
308 |
> |
|
309 |
> |
return nAtoms_ - nLocalCutoffAtoms + nCutoffGroups_; |
310 |
> |
} |
311 |
> |
|
312 |
> |
void SimInfo::calcNdfRaw() { |
313 |
> |
int ndfRaw_local; |
314 |
|
|
315 |
< |
if( !boxIsInit ) boxIsInit = 1; |
315 |
> |
MoleculeIterator i; |
316 |
> |
vector<StuntDouble*>::iterator j; |
317 |
> |
Molecule* mol; |
318 |
> |
StuntDouble* sd; |
319 |
|
|
320 |
< |
for(i=0; i < 3; i++) |
321 |
< |
for (j=0; j < 3; j++) Hmat[i][j] = theBox[i][j]; |
322 |
< |
|
323 |
< |
calcBoxL(); |
131 |
< |
calcHmatInv(); |
320 |
> |
// Raw degrees of freedom that we have to set |
321 |
> |
ndfRaw_local = 0; |
322 |
> |
|
323 |
> |
for (mol = beginMolecule(i); mol != NULL; mol = nextMolecule(i)) { |
324 |
|
|
325 |
< |
for(i=0; i < 3; i++) { |
326 |
< |
for (j=0; j < 3; j++) { |
327 |
< |
FortranHmat[3*j + i] = Hmat[i][j]; |
328 |
< |
FortranHmatInv[3*j + i] = HmatInv[i][j]; |
325 |
> |
for (sd = mol->beginIntegrableObject(j); sd != NULL; |
326 |
> |
sd = mol->nextIntegrableObject(j)) { |
327 |
> |
|
328 |
> |
ndfRaw_local += 3; |
329 |
> |
|
330 |
> |
if (sd->isDirectional()) { |
331 |
> |
if (sd->isLinear()) { |
332 |
> |
ndfRaw_local += 2; |
333 |
> |
} else { |
334 |
> |
ndfRaw_local += 3; |
335 |
> |
} |
336 |
> |
} |
337 |
> |
|
338 |
> |
} |
339 |
|
} |
340 |
+ |
|
341 |
+ |
#ifdef IS_MPI |
342 |
+ |
MPI_Allreduce(&ndfRaw_local,&ndfRaw_,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
343 |
+ |
#else |
344 |
+ |
ndfRaw_ = ndfRaw_local; |
345 |
+ |
#endif |
346 |
|
} |
347 |
|
|
348 |
< |
setFortranBox(FortranHmat, FortranHmatInv, &orthoRhombic); |
349 |
< |
|
142 |
< |
} |
143 |
< |
|
348 |
> |
void SimInfo::calcNdfTrans() { |
349 |
> |
int ndfTrans_local; |
350 |
|
|
351 |
< |
void SimInfo::getBoxM (double theBox[3][3]) { |
351 |
> |
ndfTrans_local = 3 * nIntegrableObjects_ - nConstraints_; |
352 |
|
|
147 |
– |
int i, j; |
148 |
– |
for(i=0; i<3; i++) |
149 |
– |
for (j=0; j<3; j++) theBox[i][j] = Hmat[i][j]; |
150 |
– |
} |
353 |
|
|
354 |
+ |
#ifdef IS_MPI |
355 |
+ |
MPI_Allreduce(&ndfTrans_local,&ndfTrans_,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
356 |
+ |
#else |
357 |
+ |
ndfTrans_ = ndfTrans_local; |
358 |
+ |
#endif |
359 |
|
|
360 |
< |
void SimInfo::scaleBox(double scale) { |
361 |
< |
double theBox[3][3]; |
362 |
< |
int i, j; |
360 |
> |
ndfTrans_ = ndfTrans_ - 3 - nZconstraint_; |
361 |
> |
|
362 |
> |
} |
363 |
|
|
364 |
< |
// cerr << "Scaling box by " << scale << "\n"; |
364 |
> |
void SimInfo::addInteractionPairs(Molecule* mol) { |
365 |
> |
ForceFieldOptions& options_ = forceField_->getForceFieldOptions(); |
366 |
> |
vector<Bond*>::iterator bondIter; |
367 |
> |
vector<Bend*>::iterator bendIter; |
368 |
> |
vector<Torsion*>::iterator torsionIter; |
369 |
> |
vector<Inversion*>::iterator inversionIter; |
370 |
> |
Bond* bond; |
371 |
> |
Bend* bend; |
372 |
> |
Torsion* torsion; |
373 |
> |
Inversion* inversion; |
374 |
> |
int a; |
375 |
> |
int b; |
376 |
> |
int c; |
377 |
> |
int d; |
378 |
|
|
379 |
< |
for(i=0; i<3; i++) |
380 |
< |
for (j=0; j<3; j++) theBox[i][j] = Hmat[i][j]*scale; |
379 |
> |
// atomGroups can be used to add special interaction maps between |
380 |
> |
// groups of atoms that are in two separate rigid bodies. |
381 |
> |
// However, most site-site interactions between two rigid bodies |
382 |
> |
// are probably not special, just the ones between the physically |
383 |
> |
// bonded atoms. Interactions *within* a single rigid body should |
384 |
> |
// always be excluded. These are done at the bottom of this |
385 |
> |
// function. |
386 |
|
|
387 |
< |
setBoxM(theBox); |
388 |
< |
|
389 |
< |
} |
390 |
< |
|
391 |
< |
void SimInfo::calcHmatInv( void ) { |
392 |
< |
|
393 |
< |
int oldOrtho; |
394 |
< |
int i,j; |
395 |
< |
double smallDiag; |
396 |
< |
double tol; |
397 |
< |
double sanity[3][3]; |
398 |
< |
|
399 |
< |
invertMat3( Hmat, HmatInv ); |
400 |
< |
|
401 |
< |
// check to see if Hmat is orthorhombic |
402 |
< |
|
403 |
< |
oldOrtho = orthoRhombic; |
404 |
< |
|
405 |
< |
smallDiag = fabs(Hmat[0][0]); |
406 |
< |
if(smallDiag > fabs(Hmat[1][1])) smallDiag = fabs(Hmat[1][1]); |
407 |
< |
if(smallDiag > fabs(Hmat[2][2])) smallDiag = fabs(Hmat[2][2]); |
408 |
< |
tol = smallDiag * orthoTolerance; |
409 |
< |
|
185 |
< |
orthoRhombic = 1; |
186 |
< |
|
187 |
< |
for (i = 0; i < 3; i++ ) { |
188 |
< |
for (j = 0 ; j < 3; j++) { |
189 |
< |
if (i != j) { |
190 |
< |
if (orthoRhombic) { |
191 |
< |
if ( fabs(Hmat[i][j]) >= tol) orthoRhombic = 0; |
192 |
< |
} |
387 |
> |
map<int, set<int> > atomGroups; |
388 |
> |
Molecule::RigidBodyIterator rbIter; |
389 |
> |
RigidBody* rb; |
390 |
> |
Molecule::IntegrableObjectIterator ii; |
391 |
> |
StuntDouble* sd; |
392 |
> |
|
393 |
> |
for (sd = mol->beginIntegrableObject(ii); sd != NULL; |
394 |
> |
sd = mol->nextIntegrableObject(ii)) { |
395 |
> |
|
396 |
> |
if (sd->isRigidBody()) { |
397 |
> |
rb = static_cast<RigidBody*>(sd); |
398 |
> |
vector<Atom*> atoms = rb->getAtoms(); |
399 |
> |
set<int> rigidAtoms; |
400 |
> |
for (int i = 0; i < static_cast<int>(atoms.size()); ++i) { |
401 |
> |
rigidAtoms.insert(atoms[i]->getGlobalIndex()); |
402 |
> |
} |
403 |
> |
for (int i = 0; i < static_cast<int>(atoms.size()); ++i) { |
404 |
> |
atomGroups.insert(map<int, set<int> >::value_type(atoms[i]->getGlobalIndex(), rigidAtoms)); |
405 |
> |
} |
406 |
> |
} else { |
407 |
> |
set<int> oneAtomSet; |
408 |
> |
oneAtomSet.insert(sd->getGlobalIndex()); |
409 |
> |
atomGroups.insert(map<int, set<int> >::value_type(sd->getGlobalIndex(), oneAtomSet)); |
410 |
|
} |
411 |
< |
} |
412 |
< |
} |
411 |
> |
} |
412 |
> |
|
413 |
> |
for (bond= mol->beginBond(bondIter); bond != NULL; |
414 |
> |
bond = mol->nextBond(bondIter)) { |
415 |
|
|
416 |
< |
if( oldOrtho != orthoRhombic ){ |
416 |
> |
a = bond->getAtomA()->getGlobalIndex(); |
417 |
> |
b = bond->getAtomB()->getGlobalIndex(); |
418 |
|
|
419 |
< |
if( orthoRhombic ) { |
420 |
< |
sprintf( painCave.errMsg, |
421 |
< |
"OOPSE is switching from the default Non-Orthorhombic\n" |
422 |
< |
"\tto the faster Orthorhombic periodic boundary computations.\n" |
423 |
< |
"\tThis is usually a good thing, but if you wan't the\n" |
204 |
< |
"\tNon-Orthorhombic computations, make the orthoBoxTolerance\n" |
205 |
< |
"\tvariable ( currently set to %G ) smaller.\n", |
206 |
< |
orthoTolerance); |
207 |
< |
painCave.severity = OOPSE_INFO; |
208 |
< |
simError(); |
419 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
420 |
> |
oneTwoInteractions_.addPair(a, b); |
421 |
> |
} else { |
422 |
> |
excludedInteractions_.addPair(a, b); |
423 |
> |
} |
424 |
|
} |
210 |
– |
else { |
211 |
– |
sprintf( painCave.errMsg, |
212 |
– |
"OOPSE is switching from the faster Orthorhombic to the more\n" |
213 |
– |
"\tflexible Non-Orthorhombic periodic boundary computations.\n" |
214 |
– |
"\tThis is usually because the box has deformed under\n" |
215 |
– |
"\tNPTf integration. If you wan't to live on the edge with\n" |
216 |
– |
"\tthe Orthorhombic computations, make the orthoBoxTolerance\n" |
217 |
– |
"\tvariable ( currently set to %G ) larger.\n", |
218 |
– |
orthoTolerance); |
219 |
– |
painCave.severity = OOPSE_WARNING; |
220 |
– |
simError(); |
221 |
– |
} |
222 |
– |
} |
223 |
– |
} |
425 |
|
|
426 |
< |
void SimInfo::calcBoxL( void ){ |
426 |
> |
for (bend= mol->beginBend(bendIter); bend != NULL; |
427 |
> |
bend = mol->nextBend(bendIter)) { |
428 |
|
|
429 |
< |
double dx, dy, dz, dsq; |
429 |
> |
a = bend->getAtomA()->getGlobalIndex(); |
430 |
> |
b = bend->getAtomB()->getGlobalIndex(); |
431 |
> |
c = bend->getAtomC()->getGlobalIndex(); |
432 |
> |
|
433 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
434 |
> |
oneTwoInteractions_.addPair(a, b); |
435 |
> |
oneTwoInteractions_.addPair(b, c); |
436 |
> |
} else { |
437 |
> |
excludedInteractions_.addPair(a, b); |
438 |
> |
excludedInteractions_.addPair(b, c); |
439 |
> |
} |
440 |
|
|
441 |
< |
// boxVol = Determinant of Hmat |
441 |
> |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
442 |
> |
oneThreeInteractions_.addPair(a, c); |
443 |
> |
} else { |
444 |
> |
excludedInteractions_.addPair(a, c); |
445 |
> |
} |
446 |
> |
} |
447 |
|
|
448 |
< |
boxVol = matDet3( Hmat ); |
448 |
> |
for (torsion= mol->beginTorsion(torsionIter); torsion != NULL; |
449 |
> |
torsion = mol->nextTorsion(torsionIter)) { |
450 |
|
|
451 |
< |
// boxLx |
452 |
< |
|
453 |
< |
dx = Hmat[0][0]; dy = Hmat[1][0]; dz = Hmat[2][0]; |
454 |
< |
dsq = dx*dx + dy*dy + dz*dz; |
237 |
< |
boxL[0] = sqrt( dsq ); |
238 |
< |
//maxCutoff = 0.5 * boxL[0]; |
451 |
> |
a = torsion->getAtomA()->getGlobalIndex(); |
452 |
> |
b = torsion->getAtomB()->getGlobalIndex(); |
453 |
> |
c = torsion->getAtomC()->getGlobalIndex(); |
454 |
> |
d = torsion->getAtomD()->getGlobalIndex(); |
455 |
|
|
456 |
< |
// boxLy |
457 |
< |
|
458 |
< |
dx = Hmat[0][1]; dy = Hmat[1][1]; dz = Hmat[2][1]; |
459 |
< |
dsq = dx*dx + dy*dy + dz*dz; |
460 |
< |
boxL[1] = sqrt( dsq ); |
461 |
< |
//if( (0.5 * boxL[1]) < maxCutoff ) maxCutoff = 0.5 * boxL[1]; |
456 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
457 |
> |
oneTwoInteractions_.addPair(a, b); |
458 |
> |
oneTwoInteractions_.addPair(b, c); |
459 |
> |
oneTwoInteractions_.addPair(c, d); |
460 |
> |
} else { |
461 |
> |
excludedInteractions_.addPair(a, b); |
462 |
> |
excludedInteractions_.addPair(b, c); |
463 |
> |
excludedInteractions_.addPair(c, d); |
464 |
> |
} |
465 |
|
|
466 |
+ |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
467 |
+ |
oneThreeInteractions_.addPair(a, c); |
468 |
+ |
oneThreeInteractions_.addPair(b, d); |
469 |
+ |
} else { |
470 |
+ |
excludedInteractions_.addPair(a, c); |
471 |
+ |
excludedInteractions_.addPair(b, d); |
472 |
+ |
} |
473 |
|
|
474 |
< |
// boxLz |
475 |
< |
|
476 |
< |
dx = Hmat[0][2]; dy = Hmat[1][2]; dz = Hmat[2][2]; |
477 |
< |
dsq = dx*dx + dy*dy + dz*dz; |
478 |
< |
boxL[2] = sqrt( dsq ); |
479 |
< |
//if( (0.5 * boxL[2]) < maxCutoff ) maxCutoff = 0.5 * boxL[2]; |
474 |
> |
if (options_.havevdw14scale() || options_.haveelectrostatic14scale()) { |
475 |
> |
oneFourInteractions_.addPair(a, d); |
476 |
> |
} else { |
477 |
> |
excludedInteractions_.addPair(a, d); |
478 |
> |
} |
479 |
> |
} |
480 |
|
|
481 |
< |
//calculate the max cutoff |
482 |
< |
maxCutoff = calcMaxCutOff(); |
257 |
< |
|
258 |
< |
checkCutOffs(); |
481 |
> |
for (inversion= mol->beginInversion(inversionIter); inversion != NULL; |
482 |
> |
inversion = mol->nextInversion(inversionIter)) { |
483 |
|
|
484 |
< |
} |
484 |
> |
a = inversion->getAtomA()->getGlobalIndex(); |
485 |
> |
b = inversion->getAtomB()->getGlobalIndex(); |
486 |
> |
c = inversion->getAtomC()->getGlobalIndex(); |
487 |
> |
d = inversion->getAtomD()->getGlobalIndex(); |
488 |
|
|
489 |
+ |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
490 |
+ |
oneTwoInteractions_.addPair(a, b); |
491 |
+ |
oneTwoInteractions_.addPair(a, c); |
492 |
+ |
oneTwoInteractions_.addPair(a, d); |
493 |
+ |
} else { |
494 |
+ |
excludedInteractions_.addPair(a, b); |
495 |
+ |
excludedInteractions_.addPair(a, c); |
496 |
+ |
excludedInteractions_.addPair(a, d); |
497 |
+ |
} |
498 |
|
|
499 |
< |
double SimInfo::calcMaxCutOff(){ |
499 |
> |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
500 |
> |
oneThreeInteractions_.addPair(b, c); |
501 |
> |
oneThreeInteractions_.addPair(b, d); |
502 |
> |
oneThreeInteractions_.addPair(c, d); |
503 |
> |
} else { |
504 |
> |
excludedInteractions_.addPair(b, c); |
505 |
> |
excludedInteractions_.addPair(b, d); |
506 |
> |
excludedInteractions_.addPair(c, d); |
507 |
> |
} |
508 |
> |
} |
509 |
|
|
510 |
< |
double ri[3], rj[3], rk[3]; |
511 |
< |
double rij[3], rjk[3], rki[3]; |
512 |
< |
double minDist; |
510 |
> |
for (rb = mol->beginRigidBody(rbIter); rb != NULL; |
511 |
> |
rb = mol->nextRigidBody(rbIter)) { |
512 |
> |
vector<Atom*> atoms = rb->getAtoms(); |
513 |
> |
for (int i = 0; i < static_cast<int>(atoms.size()) -1 ; ++i) { |
514 |
> |
for (int j = i + 1; j < static_cast<int>(atoms.size()); ++j) { |
515 |
> |
a = atoms[i]->getGlobalIndex(); |
516 |
> |
b = atoms[j]->getGlobalIndex(); |
517 |
> |
excludedInteractions_.addPair(a, b); |
518 |
> |
} |
519 |
> |
} |
520 |
> |
} |
521 |
|
|
522 |
< |
ri[0] = Hmat[0][0]; |
270 |
< |
ri[1] = Hmat[1][0]; |
271 |
< |
ri[2] = Hmat[2][0]; |
522 |
> |
} |
523 |
|
|
524 |
< |
rj[0] = Hmat[0][1]; |
525 |
< |
rj[1] = Hmat[1][1]; |
526 |
< |
rj[2] = Hmat[2][1]; |
524 |
> |
void SimInfo::removeInteractionPairs(Molecule* mol) { |
525 |
> |
ForceFieldOptions& options_ = forceField_->getForceFieldOptions(); |
526 |
> |
vector<Bond*>::iterator bondIter; |
527 |
> |
vector<Bend*>::iterator bendIter; |
528 |
> |
vector<Torsion*>::iterator torsionIter; |
529 |
> |
vector<Inversion*>::iterator inversionIter; |
530 |
> |
Bond* bond; |
531 |
> |
Bend* bend; |
532 |
> |
Torsion* torsion; |
533 |
> |
Inversion* inversion; |
534 |
> |
int a; |
535 |
> |
int b; |
536 |
> |
int c; |
537 |
> |
int d; |
538 |
|
|
539 |
< |
rk[0] = Hmat[0][2]; |
540 |
< |
rk[1] = Hmat[1][2]; |
541 |
< |
rk[2] = Hmat[2][2]; |
539 |
> |
map<int, set<int> > atomGroups; |
540 |
> |
Molecule::RigidBodyIterator rbIter; |
541 |
> |
RigidBody* rb; |
542 |
> |
Molecule::IntegrableObjectIterator ii; |
543 |
> |
StuntDouble* sd; |
544 |
|
|
545 |
< |
crossProduct3(ri, rj, rij); |
546 |
< |
distXY = dotProduct3(rk,rij) / norm3(rij); |
545 |
> |
for (sd = mol->beginIntegrableObject(ii); sd != NULL; |
546 |
> |
sd = mol->nextIntegrableObject(ii)) { |
547 |
> |
|
548 |
> |
if (sd->isRigidBody()) { |
549 |
> |
rb = static_cast<RigidBody*>(sd); |
550 |
> |
vector<Atom*> atoms = rb->getAtoms(); |
551 |
> |
set<int> rigidAtoms; |
552 |
> |
for (int i = 0; i < static_cast<int>(atoms.size()); ++i) { |
553 |
> |
rigidAtoms.insert(atoms[i]->getGlobalIndex()); |
554 |
> |
} |
555 |
> |
for (int i = 0; i < static_cast<int>(atoms.size()); ++i) { |
556 |
> |
atomGroups.insert(map<int, set<int> >::value_type(atoms[i]->getGlobalIndex(), rigidAtoms)); |
557 |
> |
} |
558 |
> |
} else { |
559 |
> |
set<int> oneAtomSet; |
560 |
> |
oneAtomSet.insert(sd->getGlobalIndex()); |
561 |
> |
atomGroups.insert(map<int, set<int> >::value_type(sd->getGlobalIndex(), oneAtomSet)); |
562 |
> |
} |
563 |
> |
} |
564 |
|
|
565 |
< |
crossProduct3(rj,rk, rjk); |
566 |
< |
distYZ = dotProduct3(ri,rjk) / norm3(rjk); |
565 |
> |
for (bond= mol->beginBond(bondIter); bond != NULL; |
566 |
> |
bond = mol->nextBond(bondIter)) { |
567 |
> |
|
568 |
> |
a = bond->getAtomA()->getGlobalIndex(); |
569 |
> |
b = bond->getAtomB()->getGlobalIndex(); |
570 |
> |
|
571 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
572 |
> |
oneTwoInteractions_.removePair(a, b); |
573 |
> |
} else { |
574 |
> |
excludedInteractions_.removePair(a, b); |
575 |
> |
} |
576 |
> |
} |
577 |
|
|
578 |
< |
crossProduct3(rk,ri, rki); |
579 |
< |
distZX = dotProduct3(rj,rki) / norm3(rki); |
578 |
> |
for (bend= mol->beginBend(bendIter); bend != NULL; |
579 |
> |
bend = mol->nextBend(bendIter)) { |
580 |
|
|
581 |
< |
minDist = min(min(distXY, distYZ), distZX); |
582 |
< |
return minDist/2; |
583 |
< |
|
584 |
< |
} |
581 |
> |
a = bend->getAtomA()->getGlobalIndex(); |
582 |
> |
b = bend->getAtomB()->getGlobalIndex(); |
583 |
> |
c = bend->getAtomC()->getGlobalIndex(); |
584 |
> |
|
585 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
586 |
> |
oneTwoInteractions_.removePair(a, b); |
587 |
> |
oneTwoInteractions_.removePair(b, c); |
588 |
> |
} else { |
589 |
> |
excludedInteractions_.removePair(a, b); |
590 |
> |
excludedInteractions_.removePair(b, c); |
591 |
> |
} |
592 |
|
|
593 |
< |
void SimInfo::wrapVector( double thePos[3] ){ |
594 |
< |
|
595 |
< |
int i; |
596 |
< |
double scaled[3]; |
593 |
> |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
594 |
> |
oneThreeInteractions_.removePair(a, c); |
595 |
> |
} else { |
596 |
> |
excludedInteractions_.removePair(a, c); |
597 |
> |
} |
598 |
> |
} |
599 |
|
|
600 |
< |
if( !orthoRhombic ){ |
601 |
< |
// calc the scaled coordinates. |
600 |
> |
for (torsion= mol->beginTorsion(torsionIter); torsion != NULL; |
601 |
> |
torsion = mol->nextTorsion(torsionIter)) { |
602 |
> |
|
603 |
> |
a = torsion->getAtomA()->getGlobalIndex(); |
604 |
> |
b = torsion->getAtomB()->getGlobalIndex(); |
605 |
> |
c = torsion->getAtomC()->getGlobalIndex(); |
606 |
> |
d = torsion->getAtomD()->getGlobalIndex(); |
607 |
|
|
608 |
+ |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
609 |
+ |
oneTwoInteractions_.removePair(a, b); |
610 |
+ |
oneTwoInteractions_.removePair(b, c); |
611 |
+ |
oneTwoInteractions_.removePair(c, d); |
612 |
+ |
} else { |
613 |
+ |
excludedInteractions_.removePair(a, b); |
614 |
+ |
excludedInteractions_.removePair(b, c); |
615 |
+ |
excludedInteractions_.removePair(c, d); |
616 |
+ |
} |
617 |
|
|
618 |
< |
matVecMul3(HmatInv, thePos, scaled); |
619 |
< |
|
620 |
< |
for(i=0; i<3; i++) |
621 |
< |
scaled[i] -= roundMe(scaled[i]); |
622 |
< |
|
623 |
< |
// calc the wrapped real coordinates from the wrapped scaled coordinates |
624 |
< |
|
311 |
< |
matVecMul3(Hmat, scaled, thePos); |
618 |
> |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
619 |
> |
oneThreeInteractions_.removePair(a, c); |
620 |
> |
oneThreeInteractions_.removePair(b, d); |
621 |
> |
} else { |
622 |
> |
excludedInteractions_.removePair(a, c); |
623 |
> |
excludedInteractions_.removePair(b, d); |
624 |
> |
} |
625 |
|
|
626 |
< |
} |
627 |
< |
else{ |
628 |
< |
// calc the scaled coordinates. |
629 |
< |
|
630 |
< |
for(i=0; i<3; i++) |
631 |
< |
scaled[i] = thePos[i]*HmatInv[i][i]; |
319 |
< |
|
320 |
< |
// wrap the scaled coordinates |
321 |
< |
|
322 |
< |
for(i=0; i<3; i++) |
323 |
< |
scaled[i] -= roundMe(scaled[i]); |
324 |
< |
|
325 |
< |
// calc the wrapped real coordinates from the wrapped scaled coordinates |
326 |
< |
|
327 |
< |
for(i=0; i<3; i++) |
328 |
< |
thePos[i] = scaled[i]*Hmat[i][i]; |
329 |
< |
} |
330 |
< |
|
331 |
< |
} |
626 |
> |
if (options_.havevdw14scale() || options_.haveelectrostatic14scale()) { |
627 |
> |
oneFourInteractions_.removePair(a, d); |
628 |
> |
} else { |
629 |
> |
excludedInteractions_.removePair(a, d); |
630 |
> |
} |
631 |
> |
} |
632 |
|
|
633 |
+ |
for (inversion= mol->beginInversion(inversionIter); inversion != NULL; |
634 |
+ |
inversion = mol->nextInversion(inversionIter)) { |
635 |
|
|
636 |
< |
int SimInfo::getNDF(){ |
637 |
< |
int ndf_local; |
636 |
> |
a = inversion->getAtomA()->getGlobalIndex(); |
637 |
> |
b = inversion->getAtomB()->getGlobalIndex(); |
638 |
> |
c = inversion->getAtomC()->getGlobalIndex(); |
639 |
> |
d = inversion->getAtomD()->getGlobalIndex(); |
640 |
|
|
641 |
< |
ndf_local = 0; |
642 |
< |
|
643 |
< |
for(int i = 0; i < integrableObjects.size(); i++){ |
644 |
< |
ndf_local += 3; |
645 |
< |
if (integrableObjects[i]->isDirectional()) { |
646 |
< |
if (integrableObjects[i]->isLinear()) |
647 |
< |
ndf_local += 2; |
648 |
< |
else |
649 |
< |
ndf_local += 3; |
641 |
> |
if (options_.havevdw12scale() || options_.haveelectrostatic12scale()) { |
642 |
> |
oneTwoInteractions_.removePair(a, b); |
643 |
> |
oneTwoInteractions_.removePair(a, c); |
644 |
> |
oneTwoInteractions_.removePair(a, d); |
645 |
> |
} else { |
646 |
> |
excludedInteractions_.removePair(a, b); |
647 |
> |
excludedInteractions_.removePair(a, c); |
648 |
> |
excludedInteractions_.removePair(a, d); |
649 |
> |
} |
650 |
> |
|
651 |
> |
if (options_.havevdw13scale() || options_.haveelectrostatic13scale()) { |
652 |
> |
oneThreeInteractions_.removePair(b, c); |
653 |
> |
oneThreeInteractions_.removePair(b, d); |
654 |
> |
oneThreeInteractions_.removePair(c, d); |
655 |
> |
} else { |
656 |
> |
excludedInteractions_.removePair(b, c); |
657 |
> |
excludedInteractions_.removePair(b, d); |
658 |
> |
excludedInteractions_.removePair(c, d); |
659 |
> |
} |
660 |
|
} |
661 |
+ |
|
662 |
+ |
for (rb = mol->beginRigidBody(rbIter); rb != NULL; |
663 |
+ |
rb = mol->nextRigidBody(rbIter)) { |
664 |
+ |
vector<Atom*> atoms = rb->getAtoms(); |
665 |
+ |
for (int i = 0; i < static_cast<int>(atoms.size()) -1 ; ++i) { |
666 |
+ |
for (int j = i + 1; j < static_cast<int>(atoms.size()); ++j) { |
667 |
+ |
a = atoms[i]->getGlobalIndex(); |
668 |
+ |
b = atoms[j]->getGlobalIndex(); |
669 |
+ |
excludedInteractions_.removePair(a, b); |
670 |
+ |
} |
671 |
+ |
} |
672 |
+ |
} |
673 |
+ |
|
674 |
|
} |
675 |
+ |
|
676 |
+ |
|
677 |
+ |
void SimInfo::addMoleculeStamp(MoleculeStamp* molStamp, int nmol) { |
678 |
+ |
int curStampId; |
679 |
+ |
|
680 |
+ |
//index from 0 |
681 |
+ |
curStampId = moleculeStamps_.size(); |
682 |
|
|
683 |
< |
// n_constraints is local, so subtract them on each processor: |
683 |
> |
moleculeStamps_.push_back(molStamp); |
684 |
> |
molStampIds_.insert(molStampIds_.end(), nmol, curStampId); |
685 |
> |
} |
686 |
|
|
351 |
– |
ndf_local -= n_constraints; |
687 |
|
|
688 |
+ |
/** |
689 |
+ |
* update |
690 |
+ |
* |
691 |
+ |
* Performs the global checks and variable settings after the |
692 |
+ |
* objects have been created. |
693 |
+ |
* |
694 |
+ |
*/ |
695 |
+ |
void SimInfo::update() { |
696 |
+ |
setupSimVariables(); |
697 |
+ |
calcNdf(); |
698 |
+ |
calcNdfRaw(); |
699 |
+ |
calcNdfTrans(); |
700 |
+ |
} |
701 |
+ |
|
702 |
+ |
/** |
703 |
+ |
* getSimulatedAtomTypes |
704 |
+ |
* |
705 |
+ |
* Returns an STL set of AtomType* that are actually present in this |
706 |
+ |
* simulation. Must query all processors to assemble this information. |
707 |
+ |
* |
708 |
+ |
*/ |
709 |
+ |
set<AtomType*> SimInfo::getSimulatedAtomTypes() { |
710 |
+ |
SimInfo::MoleculeIterator mi; |
711 |
+ |
Molecule* mol; |
712 |
+ |
Molecule::AtomIterator ai; |
713 |
+ |
Atom* atom; |
714 |
+ |
set<AtomType*> atomTypes; |
715 |
+ |
|
716 |
+ |
for(mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
717 |
+ |
for(atom = mol->beginAtom(ai); atom != NULL; |
718 |
+ |
atom = mol->nextAtom(ai)) { |
719 |
+ |
atomTypes.insert(atom->getAtomType()); |
720 |
+ |
} |
721 |
+ |
} |
722 |
+ |
|
723 |
|
#ifdef IS_MPI |
354 |
– |
MPI_Allreduce(&ndf_local,&ndf,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
355 |
– |
#else |
356 |
– |
ndf = ndf_local; |
357 |
– |
#endif |
724 |
|
|
725 |
< |
// nZconstraints is global, as are the 3 COM translations for the |
726 |
< |
// entire system: |
725 |
> |
// loop over the found atom types on this processor, and add their |
726 |
> |
// numerical idents to a vector: |
727 |
> |
|
728 |
> |
vector<int> foundTypes; |
729 |
> |
set<AtomType*>::iterator i; |
730 |
> |
for (i = atomTypes.begin(); i != atomTypes.end(); ++i) |
731 |
> |
foundTypes.push_back( (*i)->getIdent() ); |
732 |
|
|
733 |
< |
ndf = ndf - 3 - nZconstraints; |
733 |
> |
// count_local holds the number of found types on this processor |
734 |
> |
int count_local = foundTypes.size(); |
735 |
|
|
736 |
< |
return ndf; |
365 |
< |
} |
736 |
> |
int nproc = MPI::COMM_WORLD.Get_size(); |
737 |
|
|
738 |
< |
int SimInfo::getNDFraw() { |
739 |
< |
int ndfRaw_local; |
738 |
> |
// we need arrays to hold the counts and displacement vectors for |
739 |
> |
// all processors |
740 |
> |
vector<int> counts(nproc, 0); |
741 |
> |
vector<int> disps(nproc, 0); |
742 |
|
|
743 |
< |
// Raw degrees of freedom that we have to set |
744 |
< |
ndfRaw_local = 0; |
745 |
< |
|
746 |
< |
for(int i = 0; i < integrableObjects.size(); i++){ |
747 |
< |
ndfRaw_local += 3; |
748 |
< |
if (integrableObjects[i]->isDirectional()) { |
749 |
< |
if (integrableObjects[i]->isLinear()) |
750 |
< |
ndfRaw_local += 2; |
751 |
< |
else |
752 |
< |
ndfRaw_local += 3; |
743 |
> |
// fill the counts array |
744 |
> |
MPI::COMM_WORLD.Allgather(&count_local, 1, MPI::INT, &counts[0], |
745 |
> |
1, MPI::INT); |
746 |
> |
|
747 |
> |
// use the processor counts to compute the displacement array |
748 |
> |
disps[0] = 0; |
749 |
> |
int totalCount = counts[0]; |
750 |
> |
for (int iproc = 1; iproc < nproc; iproc++) { |
751 |
> |
disps[iproc] = disps[iproc-1] + counts[iproc-1]; |
752 |
> |
totalCount += counts[iproc]; |
753 |
|
} |
754 |
< |
} |
754 |
> |
|
755 |
> |
// we need a (possibly redundant) set of all found types: |
756 |
> |
vector<int> ftGlobal(totalCount); |
757 |
|
|
758 |
< |
#ifdef IS_MPI |
759 |
< |
MPI_Allreduce(&ndfRaw_local,&ndfRaw,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
760 |
< |
#else |
761 |
< |
ndfRaw = ndfRaw_local; |
387 |
< |
#endif |
758 |
> |
// now spray out the foundTypes to all the other processors: |
759 |
> |
MPI::COMM_WORLD.Allgatherv(&foundTypes[0], count_local, MPI::INT, |
760 |
> |
&ftGlobal[0], &counts[0], &disps[0], |
761 |
> |
MPI::INT); |
762 |
|
|
763 |
< |
return ndfRaw; |
390 |
< |
} |
763 |
> |
vector<int>::iterator j; |
764 |
|
|
765 |
< |
int SimInfo::getNDFtranslational() { |
766 |
< |
int ndfTrans_local; |
765 |
> |
// foundIdents is a stl set, so inserting an already found ident |
766 |
> |
// will have no effect. |
767 |
> |
set<int> foundIdents; |
768 |
|
|
769 |
< |
ndfTrans_local = 3 * integrableObjects.size() - n_constraints; |
770 |
< |
|
771 |
< |
|
772 |
< |
#ifdef IS_MPI |
773 |
< |
MPI_Allreduce(&ndfTrans_local,&ndfTrans,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
774 |
< |
#else |
775 |
< |
ndfTrans = ndfTrans_local; |
769 |
> |
for (j = ftGlobal.begin(); j != ftGlobal.end(); ++j) |
770 |
> |
foundIdents.insert((*j)); |
771 |
> |
|
772 |
> |
// now iterate over the foundIdents and get the actual atom types |
773 |
> |
// that correspond to these: |
774 |
> |
set<int>::iterator it; |
775 |
> |
for (it = foundIdents.begin(); it != foundIdents.end(); ++it) |
776 |
> |
atomTypes.insert( forceField_->getAtomType((*it)) ); |
777 |
> |
|
778 |
|
#endif |
779 |
|
|
780 |
< |
ndfTrans = ndfTrans - 3 - nZconstraints; |
780 |
> |
return atomTypes; |
781 |
> |
} |
782 |
|
|
783 |
< |
return ndfTrans; |
784 |
< |
} |
783 |
> |
void SimInfo::setupSimVariables() { |
784 |
> |
useAtomicVirial_ = simParams_->getUseAtomicVirial(); |
785 |
> |
// we only call setAccumulateBoxDipole if the accumulateBoxDipole |
786 |
> |
// parameter is true |
787 |
> |
calcBoxDipole_ = false; |
788 |
> |
if ( simParams_->haveAccumulateBoxDipole() ) |
789 |
> |
if ( simParams_->getAccumulateBoxDipole() ) { |
790 |
> |
calcBoxDipole_ = true; |
791 |
> |
} |
792 |
> |
|
793 |
> |
set<AtomType*>::iterator i; |
794 |
> |
set<AtomType*> atomTypes; |
795 |
> |
atomTypes = getSimulatedAtomTypes(); |
796 |
> |
bool usesElectrostatic = false; |
797 |
> |
bool usesMetallic = false; |
798 |
> |
bool usesDirectional = false; |
799 |
> |
bool usesFluctuatingCharges = false; |
800 |
> |
//loop over all of the atom types |
801 |
> |
for (i = atomTypes.begin(); i != atomTypes.end(); ++i) { |
802 |
> |
usesElectrostatic |= (*i)->isElectrostatic(); |
803 |
> |
usesMetallic |= (*i)->isMetal(); |
804 |
> |
usesDirectional |= (*i)->isDirectional(); |
805 |
> |
usesFluctuatingCharges |= (*i)->isFluctuatingCharge(); |
806 |
> |
} |
807 |
|
|
808 |
< |
int SimInfo::getTotIntegrableObjects() { |
809 |
< |
int nObjs_local; |
810 |
< |
int nObjs; |
808 |
> |
#ifdef IS_MPI |
809 |
> |
bool temp; |
810 |
> |
temp = usesDirectional; |
811 |
> |
MPI::COMM_WORLD.Allreduce(&temp, &usesDirectionalAtoms_, 1, MPI::BOOL, |
812 |
> |
MPI::LOR); |
813 |
> |
|
814 |
> |
temp = usesMetallic; |
815 |
> |
MPI::COMM_WORLD.Allreduce(&temp, &usesMetallicAtoms_, 1, MPI::BOOL, |
816 |
> |
MPI::LOR); |
817 |
> |
|
818 |
> |
temp = usesElectrostatic; |
819 |
> |
MPI::COMM_WORLD.Allreduce(&temp, &usesElectrostaticAtoms_, 1, MPI::BOOL, |
820 |
> |
MPI::LOR); |
821 |
|
|
822 |
< |
nObjs_local = integrableObjects.size(); |
822 |
> |
temp = usesFluctuatingCharges; |
823 |
> |
MPI::COMM_WORLD.Allreduce(&temp, &usesFluctuatingCharges_, 1, MPI::BOOL, |
824 |
> |
MPI::LOR); |
825 |
> |
#else |
826 |
|
|
827 |
+ |
usesDirectionalAtoms_ = usesDirectional; |
828 |
+ |
usesMetallicAtoms_ = usesMetallic; |
829 |
+ |
usesElectrostaticAtoms_ = usesElectrostatic; |
830 |
+ |
usesFluctuatingCharges_ = usesFluctuatingCharges; |
831 |
|
|
416 |
– |
#ifdef IS_MPI |
417 |
– |
MPI_Allreduce(&nObjs_local,&nObjs,1,MPI_INT,MPI_SUM, MPI_COMM_WORLD); |
418 |
– |
#else |
419 |
– |
nObjs = nObjs_local; |
832 |
|
#endif |
833 |
+ |
|
834 |
+ |
requiresPrepair_ = usesMetallicAtoms_ ? true : false; |
835 |
+ |
requiresSkipCorrection_ = usesElectrostaticAtoms_ ? true : false; |
836 |
+ |
requiresSelfCorrection_ = usesElectrostaticAtoms_ ? true : false; |
837 |
+ |
} |
838 |
|
|
839 |
|
|
840 |
< |
return nObjs; |
841 |
< |
} |
840 |
> |
vector<int> SimInfo::getGlobalAtomIndices() { |
841 |
> |
SimInfo::MoleculeIterator mi; |
842 |
> |
Molecule* mol; |
843 |
> |
Molecule::AtomIterator ai; |
844 |
> |
Atom* atom; |
845 |
|
|
846 |
< |
void SimInfo::refreshSim(){ |
846 |
> |
vector<int> GlobalAtomIndices(getNAtoms(), 0); |
847 |
> |
|
848 |
> |
for (mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
849 |
> |
|
850 |
> |
for (atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) { |
851 |
> |
GlobalAtomIndices[atom->getLocalIndex()] = atom->getGlobalIndex(); |
852 |
> |
} |
853 |
> |
} |
854 |
> |
return GlobalAtomIndices; |
855 |
> |
} |
856 |
|
|
428 |
– |
simtype fInfo; |
429 |
– |
int isError; |
430 |
– |
int n_global; |
431 |
– |
int* excl; |
857 |
|
|
858 |
< |
fInfo.dielect = 0.0; |
858 |
> |
vector<int> SimInfo::getGlobalGroupIndices() { |
859 |
> |
SimInfo::MoleculeIterator mi; |
860 |
> |
Molecule* mol; |
861 |
> |
Molecule::CutoffGroupIterator ci; |
862 |
> |
CutoffGroup* cg; |
863 |
|
|
864 |
< |
if( useDipoles ){ |
865 |
< |
if( useReactionField )fInfo.dielect = dielectric; |
864 |
> |
vector<int> GlobalGroupIndices; |
865 |
> |
|
866 |
> |
for (mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
867 |
> |
|
868 |
> |
//local index of cutoff group is trivial, it only depends on the |
869 |
> |
//order of travesing |
870 |
> |
for (cg = mol->beginCutoffGroup(ci); cg != NULL; |
871 |
> |
cg = mol->nextCutoffGroup(ci)) { |
872 |
> |
GlobalGroupIndices.push_back(cg->getGlobalIndex()); |
873 |
> |
} |
874 |
> |
} |
875 |
> |
return GlobalGroupIndices; |
876 |
|
} |
877 |
|
|
439 |
– |
fInfo.SIM_uses_PBC = usePBC; |
878 |
|
|
879 |
< |
if (useSticky || useDipoles || useGayBerne || useShapes) { |
880 |
< |
useDirectionalAtoms = 1; |
443 |
< |
fInfo.SIM_uses_DirectionalAtoms = useDirectionalAtoms; |
444 |
< |
} |
879 |
> |
void SimInfo::prepareTopology() { |
880 |
> |
int nExclude, nOneTwo, nOneThree, nOneFour; |
881 |
|
|
882 |
< |
fInfo.SIM_uses_LennardJones = useLennardJones; |
882 |
> |
//calculate mass ratio of cutoff group |
883 |
> |
SimInfo::MoleculeIterator mi; |
884 |
> |
Molecule* mol; |
885 |
> |
Molecule::CutoffGroupIterator ci; |
886 |
> |
CutoffGroup* cg; |
887 |
> |
Molecule::AtomIterator ai; |
888 |
> |
Atom* atom; |
889 |
> |
RealType totalMass; |
890 |
|
|
891 |
< |
if (useCharges || useDipoles) { |
892 |
< |
useElectrostatics = 1; |
893 |
< |
fInfo.SIM_uses_Electrostatics = useElectrostatics; |
894 |
< |
} |
891 |
> |
/** |
892 |
> |
* The mass factor is the relative mass of an atom to the total |
893 |
> |
* mass of the cutoff group it belongs to. By default, all atoms |
894 |
> |
* are their own cutoff groups, and therefore have mass factors of |
895 |
> |
* 1. We need some special handling for massless atoms, which |
896 |
> |
* will be treated as carrying the entire mass of the cutoff |
897 |
> |
* group. |
898 |
> |
*/ |
899 |
> |
massFactors_.clear(); |
900 |
> |
massFactors_.resize(getNAtoms(), 1.0); |
901 |
> |
|
902 |
> |
for(mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
903 |
> |
for (cg = mol->beginCutoffGroup(ci); cg != NULL; |
904 |
> |
cg = mol->nextCutoffGroup(ci)) { |
905 |
|
|
906 |
< |
fInfo.SIM_uses_Charges = useCharges; |
907 |
< |
fInfo.SIM_uses_Dipoles = useDipoles; |
908 |
< |
fInfo.SIM_uses_Sticky = useSticky; |
909 |
< |
fInfo.SIM_uses_GayBerne = useGayBerne; |
910 |
< |
fInfo.SIM_uses_EAM = useEAM; |
911 |
< |
fInfo.SIM_uses_Shapes = useShapes; |
912 |
< |
fInfo.SIM_uses_FLARB = useFLARB; |
913 |
< |
fInfo.SIM_uses_RF = useReactionField; |
906 |
> |
totalMass = cg->getMass(); |
907 |
> |
for(atom = cg->beginAtom(ai); atom != NULL; atom = cg->nextAtom(ai)) { |
908 |
> |
// Check for massless groups - set mfact to 1 if true |
909 |
> |
if (totalMass != 0) |
910 |
> |
massFactors_[atom->getLocalIndex()] = atom->getMass()/totalMass; |
911 |
> |
else |
912 |
> |
massFactors_[atom->getLocalIndex()] = 1.0; |
913 |
> |
} |
914 |
> |
} |
915 |
> |
} |
916 |
|
|
917 |
< |
n_exclude = excludes->getSize(); |
463 |
< |
excl = excludes->getFortranArray(); |
464 |
< |
|
465 |
< |
#ifdef IS_MPI |
466 |
< |
n_global = mpiSim->getNAtomsGlobal(); |
467 |
< |
#else |
468 |
< |
n_global = n_atoms; |
469 |
< |
#endif |
470 |
< |
|
471 |
< |
isError = 0; |
472 |
< |
|
473 |
< |
getFortranGroupArrays(this, FglobalGroupMembership, mfact); |
474 |
< |
//it may not be a good idea to pass the address of first element in vector |
475 |
< |
//since c++ standard does not require vector to be stored continuously in meomory |
476 |
< |
//Most of the compilers will organize the memory of vector continuously |
477 |
< |
setFortranSim( &fInfo, &n_global, &n_atoms, identArray, &n_exclude, excl, |
478 |
< |
&nGlobalExcludes, globalExcludes, molMembershipArray, |
479 |
< |
&mfact[0], &ngroup, &FglobalGroupMembership[0], &isError); |
917 |
> |
// Build the identArray_ |
918 |
|
|
919 |
< |
if( isError ){ |
919 |
> |
identArray_.clear(); |
920 |
> |
identArray_.reserve(getNAtoms()); |
921 |
> |
for(mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
922 |
> |
for(atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) { |
923 |
> |
identArray_.push_back(atom->getIdent()); |
924 |
> |
} |
925 |
> |
} |
926 |
|
|
927 |
< |
sprintf( painCave.errMsg, |
484 |
< |
"There was an error setting the simulation information in fortran.\n" ); |
485 |
< |
painCave.isFatal = 1; |
486 |
< |
painCave.severity = OOPSE_ERROR; |
487 |
< |
simError(); |
488 |
< |
} |
489 |
< |
|
490 |
< |
#ifdef IS_MPI |
491 |
< |
sprintf( checkPointMsg, |
492 |
< |
"succesfully sent the simulation information to fortran.\n"); |
493 |
< |
MPIcheckPoint(); |
494 |
< |
#endif // is_mpi |
495 |
< |
|
496 |
< |
this->ndf = this->getNDF(); |
497 |
< |
this->ndfRaw = this->getNDFraw(); |
498 |
< |
this->ndfTrans = this->getNDFtranslational(); |
499 |
< |
} |
927 |
> |
//scan topology |
928 |
|
|
929 |
< |
void SimInfo::setDefaultRcut( double theRcut ){ |
930 |
< |
|
931 |
< |
haveRcut = 1; |
932 |
< |
rCut = theRcut; |
505 |
< |
rList = rCut + 1.0; |
506 |
< |
|
507 |
< |
notifyFortranCutoffs( &rCut, &rSw, &rList ); |
508 |
< |
} |
929 |
> |
nExclude = excludedInteractions_.getSize(); |
930 |
> |
nOneTwo = oneTwoInteractions_.getSize(); |
931 |
> |
nOneThree = oneThreeInteractions_.getSize(); |
932 |
> |
nOneFour = oneFourInteractions_.getSize(); |
933 |
|
|
934 |
< |
void SimInfo::setDefaultRcut( double theRcut, double theRsw ){ |
934 |
> |
int* excludeList = excludedInteractions_.getPairList(); |
935 |
> |
int* oneTwoList = oneTwoInteractions_.getPairList(); |
936 |
> |
int* oneThreeList = oneThreeInteractions_.getPairList(); |
937 |
> |
int* oneFourList = oneFourInteractions_.getPairList(); |
938 |
|
|
939 |
< |
rSw = theRsw; |
940 |
< |
setDefaultRcut( theRcut ); |
514 |
< |
} |
939 |
> |
topologyDone_ = true; |
940 |
> |
} |
941 |
|
|
942 |
+ |
void SimInfo::addProperty(GenericData* genData) { |
943 |
+ |
properties_.addProperty(genData); |
944 |
+ |
} |
945 |
|
|
946 |
< |
void SimInfo::checkCutOffs( void ){ |
947 |
< |
|
519 |
< |
if( boxIsInit ){ |
520 |
< |
|
521 |
< |
//we need to check cutOffs against the box |
522 |
< |
|
523 |
< |
if( rCut > maxCutoff ){ |
524 |
< |
sprintf( painCave.errMsg, |
525 |
< |
"cutoffRadius is too large for the current periodic box.\n" |
526 |
< |
"\tCurrent Value of cutoffRadius = %G at time %G\n " |
527 |
< |
"\tThis is larger than half of at least one of the\n" |
528 |
< |
"\tperiodic box vectors. Right now, the Box matrix is:\n" |
529 |
< |
"\n" |
530 |
< |
"\t[ %G %G %G ]\n" |
531 |
< |
"\t[ %G %G %G ]\n" |
532 |
< |
"\t[ %G %G %G ]\n", |
533 |
< |
rCut, currentTime, |
534 |
< |
Hmat[0][0], Hmat[0][1], Hmat[0][2], |
535 |
< |
Hmat[1][0], Hmat[1][1], Hmat[1][2], |
536 |
< |
Hmat[2][0], Hmat[2][1], Hmat[2][2]); |
537 |
< |
painCave.severity = OOPSE_ERROR; |
538 |
< |
painCave.isFatal = 1; |
539 |
< |
simError(); |
540 |
< |
} |
541 |
< |
} else { |
542 |
< |
// initialize this stuff before using it, OK? |
543 |
< |
sprintf( painCave.errMsg, |
544 |
< |
"Trying to check cutoffs without a box.\n" |
545 |
< |
"\tOOPSE should have better programmers than that.\n" ); |
546 |
< |
painCave.severity = OOPSE_ERROR; |
547 |
< |
painCave.isFatal = 1; |
548 |
< |
simError(); |
946 |
> |
void SimInfo::removeProperty(const string& propName) { |
947 |
> |
properties_.removeProperty(propName); |
948 |
|
} |
550 |
– |
|
551 |
– |
} |
949 |
|
|
950 |
< |
void SimInfo::addProperty(GenericData* prop){ |
950 |
> |
void SimInfo::clearProperties() { |
951 |
> |
properties_.clearProperties(); |
952 |
> |
} |
953 |
|
|
954 |
< |
map<string, GenericData*>::iterator result; |
955 |
< |
result = properties.find(prop->getID()); |
956 |
< |
|
558 |
< |
//we can't simply use properties[prop->getID()] = prop, |
559 |
< |
//it will cause memory leak if we already contain a propery which has the same name of prop |
560 |
< |
|
561 |
< |
if(result != properties.end()){ |
562 |
< |
|
563 |
< |
delete (*result).second; |
564 |
< |
(*result).second = prop; |
954 |
> |
vector<string> SimInfo::getPropertyNames() { |
955 |
> |
return properties_.getPropertyNames(); |
956 |
> |
} |
957 |
|
|
958 |
+ |
vector<GenericData*> SimInfo::getProperties() { |
959 |
+ |
return properties_.getProperties(); |
960 |
|
} |
567 |
– |
else{ |
961 |
|
|
962 |
< |
properties[prop->getID()] = prop; |
963 |
< |
|
962 |
> |
GenericData* SimInfo::getPropertyByName(const string& propName) { |
963 |
> |
return properties_.getPropertyByName(propName); |
964 |
|
} |
572 |
– |
|
573 |
– |
} |
965 |
|
|
966 |
< |
GenericData* SimInfo::getProperty(const string& propName){ |
966 |
> |
void SimInfo::setSnapshotManager(SnapshotManager* sman) { |
967 |
> |
if (sman_ == sman) { |
968 |
> |
return; |
969 |
> |
} |
970 |
> |
delete sman_; |
971 |
> |
sman_ = sman; |
972 |
> |
|
973 |
> |
Molecule* mol; |
974 |
> |
RigidBody* rb; |
975 |
> |
Atom* atom; |
976 |
> |
CutoffGroup* cg; |
977 |
> |
SimInfo::MoleculeIterator mi; |
978 |
> |
Molecule::RigidBodyIterator rbIter; |
979 |
> |
Molecule::AtomIterator atomIter; |
980 |
> |
Molecule::CutoffGroupIterator cgIter; |
981 |
|
|
982 |
< |
map<string, GenericData*>::iterator result; |
983 |
< |
|
984 |
< |
//string lowerCaseName = (); |
985 |
< |
|
986 |
< |
result = properties.find(propName); |
987 |
< |
|
988 |
< |
if(result != properties.end()) |
989 |
< |
return (*result).second; |
990 |
< |
else |
991 |
< |
return NULL; |
992 |
< |
} |
982 |
> |
for (mol = beginMolecule(mi); mol != NULL; mol = nextMolecule(mi)) { |
983 |
> |
|
984 |
> |
for (atom = mol->beginAtom(atomIter); atom != NULL; |
985 |
> |
atom = mol->nextAtom(atomIter)) { |
986 |
> |
atom->setSnapshotManager(sman_); |
987 |
> |
} |
988 |
> |
|
989 |
> |
for (rb = mol->beginRigidBody(rbIter); rb != NULL; |
990 |
> |
rb = mol->nextRigidBody(rbIter)) { |
991 |
> |
rb->setSnapshotManager(sman_); |
992 |
> |
} |
993 |
|
|
994 |
+ |
for (cg = mol->beginCutoffGroup(cgIter); cg != NULL; |
995 |
+ |
cg = mol->nextCutoffGroup(cgIter)) { |
996 |
+ |
cg->setSnapshotManager(sman_); |
997 |
+ |
} |
998 |
+ |
} |
999 |
+ |
|
1000 |
+ |
} |
1001 |
|
|
1002 |
< |
void SimInfo::getFortranGroupArrays(SimInfo* info, |
1003 |
< |
vector<int>& FglobalGroupMembership, |
1004 |
< |
vector<double>& mfact){ |
1002 |
> |
|
1003 |
> |
ostream& operator <<(ostream& o, SimInfo& info) { |
1004 |
> |
|
1005 |
> |
return o; |
1006 |
> |
} |
1007 |
> |
|
1008 |
|
|
1009 |
< |
Molecule* myMols; |
1010 |
< |
Atom** myAtoms; |
1011 |
< |
int numAtom; |
1012 |
< |
double mtot; |
1013 |
< |
int numMol; |
1014 |
< |
int numCutoffGroups; |
1015 |
< |
CutoffGroup* myCutoffGroup; |
1016 |
< |
vector<CutoffGroup*>::iterator iterCutoff; |
1017 |
< |
Atom* cutoffAtom; |
1018 |
< |
vector<Atom*>::iterator iterAtom; |
1019 |
< |
int atomIndex; |
605 |
< |
double totalMass; |
1009 |
> |
StuntDouble* SimInfo::getIOIndexToIntegrableObject(int index) { |
1010 |
> |
if (index >= IOIndexToIntegrableObject.size()) { |
1011 |
> |
sprintf(painCave.errMsg, |
1012 |
> |
"SimInfo::getIOIndexToIntegrableObject Error: Integrable Object\n" |
1013 |
> |
"\tindex exceeds number of known objects!\n"); |
1014 |
> |
painCave.isFatal = 1; |
1015 |
> |
simError(); |
1016 |
> |
return NULL; |
1017 |
> |
} else |
1018 |
> |
return IOIndexToIntegrableObject.at(index); |
1019 |
> |
} |
1020 |
|
|
1021 |
< |
mfact.clear(); |
1022 |
< |
FglobalGroupMembership.clear(); |
1023 |
< |
|
1021 |
> |
void SimInfo::setIOIndexToIntegrableObject(const vector<StuntDouble*>& v) { |
1022 |
> |
IOIndexToIntegrableObject= v; |
1023 |
> |
} |
1024 |
|
|
1025 |
< |
// Fix the silly fortran indexing problem |
1025 |
> |
int SimInfo::getNGlobalConstraints() { |
1026 |
> |
int nGlobalConstraints; |
1027 |
|
#ifdef IS_MPI |
1028 |
< |
numAtom = mpiSim->getNAtomsGlobal(); |
1028 |
> |
MPI_Allreduce(&nConstraints_, &nGlobalConstraints, 1, MPI_INT, MPI_SUM, |
1029 |
> |
MPI_COMM_WORLD); |
1030 |
|
#else |
1031 |
< |
numAtom = n_atoms; |
1031 |
> |
nGlobalConstraints = nConstraints_; |
1032 |
|
#endif |
1033 |
< |
for (int i = 0; i < numAtom; i++) |
618 |
< |
FglobalGroupMembership.push_back(globalGroupMembership[i] + 1); |
619 |
< |
|
620 |
< |
|
621 |
< |
myMols = info->molecules; |
622 |
< |
numMol = info->n_mol; |
623 |
< |
for(int i = 0; i < numMol; i++){ |
624 |
< |
numCutoffGroups = myMols[i].getNCutoffGroups(); |
625 |
< |
for(myCutoffGroup =myMols[i].beginCutoffGroup(iterCutoff); |
626 |
< |
myCutoffGroup != NULL; |
627 |
< |
myCutoffGroup =myMols[i].nextCutoffGroup(iterCutoff)){ |
628 |
< |
|
629 |
< |
totalMass = myCutoffGroup->getMass(); |
630 |
< |
|
631 |
< |
for(cutoffAtom = myCutoffGroup->beginAtom(iterAtom); |
632 |
< |
cutoffAtom != NULL; |
633 |
< |
cutoffAtom = myCutoffGroup->nextAtom(iterAtom)){ |
634 |
< |
mfact.push_back(cutoffAtom->getMass()/totalMass); |
635 |
< |
} |
636 |
< |
} |
1033 |
> |
return nGlobalConstraints; |
1034 |
|
} |
1035 |
|
|
1036 |
< |
} |
1036 |
> |
}//end namespace OpenMD |
1037 |
> |
|