ViewVC Help

View File | Revision Log | Show Annotations | View Changeset | Root Listing

root/OpenMD/trunk/src/rnemd/RNEMD.cpp

Comparing:
branches/development/src/integrators/RNEMD.cpp (file contents), Revision 1465 by chuckv, Fri Jul 9 23:08:25 2010 UTC vs.
trunk/src/rnemd/RNEMD.cpp (file contents), Revision 1971 by gezelter, Fri Feb 28 13:25:13 2014 UTC

#	Line 35 | Line 35
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).
38	>	* [3]  Sun, Lin & Gezelter, J. Chem. Phys. 128, 234107 (2008).
39		* [4]  Vardeman & Gezelter, in progress (2009).
40		*/
41	+	#ifdef IS_MPI
42	+	#include <mpi.h>
43	+	#endif
44
45		#include <cmath>
46	<	#include "integrators/RNEMD.hpp"
46	>	#include <sstream>
47	>	#include <string>
48	>
49	>	#include "rnemd/RNEMD.hpp"
50		#include "math/Vector3.hpp"
51	+	#include "math/Vector.hpp"
52		#include "math/SquareMatrix3.hpp"
53		#include "math/Polynomial.hpp"
54		#include "primitives/Molecule.hpp"
55		#include "primitives/StuntDouble.hpp"
56		#include "utils/PhysicalConstants.hpp"
57		#include "utils/Tuple.hpp"
58	+	#include "brains/Thermo.hpp"
59	+	#include "math/ConvexHull.hpp"
60
61	<	#ifndef IS_MPI
62	<	#include "math/SeqRandNumGen.hpp"
63	<	#else
55	<	#include "math/ParallelRandNumGen.hpp"
61	>	#ifdef _MSC_VER
62	>	#define isnan(x) _isnan((x))
63	>	#define isinf(x) (!_finite(x) && !_isnan(x))
64		#endif
65
66		#define HONKING_LARGE_VALUE 1.0e10
67
68	+	using namespace std;
69		namespace OpenMD {
70
71	<	RNEMD::RNEMD(SimInfo* info) : info_(info), evaluator_(info), seleMan_(info), usePeriodicBoundaryConditions_(info->getSimParams()->getUsePeriodicBoundaryConditions()) {
71	>	RNEMD::RNEMD(SimInfo* info) : info_(info), evaluator_(info), seleMan_(info),
72	>	evaluatorA_(info), seleManA_(info),
73	>	commonA_(info), evaluatorB_(info),
74	>	seleManB_(info), commonB_(info),
75	>	hasData_(false), hasDividingArea_(false),
76	>	usePeriodicBoundaryConditions_(info->getSimParams()->getUsePeriodicBoundaryConditions()) {
77
78	+	trialCount_ = 0;
79		failTrialCount_ = 0;
80		failRootCount_ = 0;
81
82	<	int seedValue;
83	<	Globals * simParams = info->getSimParams();
82	>	Globals* simParams = info->getSimParams();
83	>	RNEMDParameters* rnemdParams = simParams->getRNEMDParameters();
84
85	<	stringToEnumMap_["KineticSwap"] = rnemdKineticSwap;
86	<	stringToEnumMap_["KineticScale"] = rnemdKineticScale;
72	<	stringToEnumMap_["PxScale"] = rnemdPxScale;
73	<	stringToEnumMap_["PyScale"] = rnemdPyScale;
74	<	stringToEnumMap_["PzScale"] = rnemdPzScale;
75	<	stringToEnumMap_["Px"] = rnemdPx;
76	<	stringToEnumMap_["Py"] = rnemdPy;
77	<	stringToEnumMap_["Pz"] = rnemdPz;
78	<	stringToEnumMap_["Unknown"] = rnemdUnknown;
85	>	doRNEMD_ = rnemdParams->getUseRNEMD();
86	>	if (!doRNEMD_) return;
87
88	<	rnemdObjectSelection_ = simParams->getRNEMD_objectSelection();
89	<	evaluator_.loadScriptString(rnemdObjectSelection_);
90	<	seleMan_.setSelectionSet(evaluator_.evaluate());
88	>	stringToMethod_["Swap"]  = rnemdSwap;
89	>	stringToMethod_["NIVS"]  = rnemdNIVS;
90	>	stringToMethod_["VSS"]   = rnemdVSS;
91
92	<	// do some sanity checking
92	>	stringToFluxType_["KE"]  = rnemdKE;
93	>	stringToFluxType_["Px"]  = rnemdPx;
94	>	stringToFluxType_["Py"]  = rnemdPy;
95	>	stringToFluxType_["Pz"]  = rnemdPz;
96	>	stringToFluxType_["Pvector"]  = rnemdPvector;
97	>	stringToFluxType_["Lx"]  = rnemdLx;
98	>	stringToFluxType_["Ly"]  = rnemdLy;
99	>	stringToFluxType_["Lz"]  = rnemdLz;
100	>	stringToFluxType_["Lvector"]  = rnemdLvector;
101	>	stringToFluxType_["KE+Px"]  = rnemdKePx;
102	>	stringToFluxType_["KE+Py"]  = rnemdKePy;
103	>	stringToFluxType_["KE+Pvector"]  = rnemdKePvector;
104	>	stringToFluxType_["KE+Lx"]  = rnemdKeLx;
105	>	stringToFluxType_["KE+Ly"]  = rnemdKeLy;
106	>	stringToFluxType_["KE+Lz"]  = rnemdKeLz;
107	>	stringToFluxType_["KE+Lvector"]  = rnemdKeLvector;
108
109	<	int selectionCount = seleMan_.getSelectionCount();
110	<	int nIntegrable = info->getNGlobalIntegrableObjects();
109	>	runTime_ = simParams->getRunTime();
110	>	statusTime_ = simParams->getStatusTime();
111
112	<	if (selectionCount > nIntegrable) {
112	>	const string methStr = rnemdParams->getMethod();
113	>	bool hasFluxType = rnemdParams->haveFluxType();
114	>
115	>	rnemdObjectSelection_ = rnemdParams->getObjectSelection();
116	>
117	>	string fluxStr;
118	>	if (hasFluxType) {
119	>	fluxStr = rnemdParams->getFluxType();
120	>	} else {
121		sprintf(painCave.errMsg,
122	<	"RNEMD warning: The current RNEMD_objectSelection,\n"
123	<	"\t\t%s\n"
124	<	"\thas resulted in %d selected objects.  However,\n"
125	<	"\tthe total number of integrable objects in the system\n"
126	<	"\tis only %d.  This is almost certainly not what you want\n"
127	<	"\tto do.  A likely cause of this is forgetting the _RB_0\n"
128	<	"\tselector in the selection script!\n",
98	<	rnemdObjectSelection_.c_str(),
99	<	selectionCount, nIntegrable);
100	<	painCave.isFatal = 0;
122	>	"RNEMD: No fluxType was set in the md file.  This parameter,\n"
123	>	"\twhich must be one of the following values:\n"
124	>	"\tKE, Px, Py, Pz, Pvector, Lx, Ly, Lz, Lvector,\n"
125	>	"\tKE+Px, KE+Py, KE+Pvector, KE+Lx, KE+Ly, KE+Lz, KE+Lvector\n"
126	>	"\tmust be set to use RNEMD\n");
127	>	painCave.isFatal = 1;
128	>	painCave.severity = OPENMD_ERROR;
129		simError();
102	–
130		}
131	+
132	+	bool hasKineticFlux = rnemdParams->haveKineticFlux();
133	+	bool hasMomentumFlux = rnemdParams->haveMomentumFlux();
134	+	bool hasMomentumFluxVector = rnemdParams->haveMomentumFluxVector();
135	+	bool hasAngularMomentumFlux = rnemdParams->haveAngularMomentumFlux();
136	+	bool hasAngularMomentumFluxVector = rnemdParams->haveAngularMomentumFluxVector();
137	+	hasSelectionA_ = rnemdParams->haveSelectionA();
138	+	hasSelectionB_ = rnemdParams->haveSelectionB();
139	+	bool hasSlabWidth = rnemdParams->haveSlabWidth();
140	+	bool hasSlabACenter = rnemdParams->haveSlabACenter();
141	+	bool hasSlabBCenter = rnemdParams->haveSlabBCenter();
142	+	bool hasSphereARadius = rnemdParams->haveSphereARadius();
143	+	hasSphereBRadius_ = rnemdParams->haveSphereBRadius();
144	+	bool hasCoordinateOrigin = rnemdParams->haveCoordinateOrigin();
145	+	bool hasOutputFileName = rnemdParams->haveOutputFileName();
146	+	bool hasOutputFields = rnemdParams->haveOutputFields();
147
148	<	const std::string st = simParams->getRNEMD_exchangeType();
148	>	map<string, RNEMDMethod>::iterator i;
149	>	i = stringToMethod_.find(methStr);
150	>	if (i != stringToMethod_.end())
151	>	rnemdMethod_ = i->second;
152	>	else {
153	>	sprintf(painCave.errMsg,
154	>	"RNEMD: The current method,\n"
155	>	"\t\t%s is not one of the recognized\n"
156	>	"\texchange methods: Swap, NIVS, or VSS\n",
157	>	methStr.c_str());
158	>	painCave.isFatal = 1;
159	>	painCave.severity = OPENMD_ERROR;
160	>	simError();
161	>	}
162
163	<	std::map<std::string, RNEMDTypeEnum>::iterator i;
164	<	i = stringToEnumMap_.find(st);
165	<	rnemdType_ = (i == stringToEnumMap_.end()) ? RNEMD::rnemdUnknown : i->second;
166	<	if (rnemdType_ == rnemdUnknown) {
167	<	std::cerr << "WARNING! RNEMD Type Unknown!\n";
163	>	map<string, RNEMDFluxType>::iterator j;
164	>	j = stringToFluxType_.find(fluxStr);
165	>	if (j != stringToFluxType_.end())
166	>	rnemdFluxType_ = j->second;
167	>	else {
168	>	sprintf(painCave.errMsg,
169	>	"RNEMD: The current fluxType,\n"
170	>	"\t\t%s\n"
171	>	"\tis not one of the recognized flux types.\n",
172	>	fluxStr.c_str());
173	>	painCave.isFatal = 1;
174	>	painCave.severity = OPENMD_ERROR;
175	>	simError();
176		}
177
178	<	#ifdef IS_MPI
179	<	if (worldRank == 0) {
180	<	#endif
181	<
182	<	std::string rnemdFileName;
183	<	std::string xTempFileName;
184	<	std::string yTempFileName;
121	<	std::string zTempFileName;
122	<	switch(rnemdType_) {
123	<	case rnemdKineticSwap :
124	<	case rnemdKineticScale :
125	<	rnemdFileName = "temperature.log";
178	>	bool methodFluxMismatch = false;
179	>	bool hasCorrectFlux = false;
180	>	switch(rnemdMethod_) {
181	>	case rnemdSwap:
182	>	switch (rnemdFluxType_) {
183	>	case rnemdKE:
184	>	hasCorrectFlux = hasKineticFlux;
185		break;
186	<	case rnemdPx :
187	<	case rnemdPxScale :
188	<	case rnemdPy :
189	<	case rnemdPyScale :
131	<	rnemdFileName = "momemtum.log";
132	<	xTempFileName = "temperatureX.log";
133	<	yTempFileName = "temperatureY.log";
134	<	zTempFileName = "temperatureZ.log";
135	<	xTempLog_.open(xTempFileName.c_str());
136	<	yTempLog_.open(yTempFileName.c_str());
137	<	zTempLog_.open(zTempFileName.c_str());
186	>	case rnemdPx:
187	>	case rnemdPy:
188	>	case rnemdPz:
189	>	hasCorrectFlux = hasMomentumFlux;
190		break;
139	–	case rnemdPz :
140	–	case rnemdPzScale :
141	–	case rnemdUnknown :
191		default :
192	<	rnemdFileName = "rnemd.log";
192	>	methodFluxMismatch = true;
193		break;
194		}
195	<	rnemdLog_.open(rnemdFileName.c_str());
196	<
197	<	#ifdef IS_MPI
195	>	break;
196	>	case rnemdNIVS:
197	>	switch (rnemdFluxType_) {
198	>	case rnemdKE:
199	>	case rnemdRotKE:
200	>	case rnemdFullKE:
201	>	hasCorrectFlux = hasKineticFlux;
202	>	break;
203	>	case rnemdPx:
204	>	case rnemdPy:
205	>	case rnemdPz:
206	>	hasCorrectFlux = hasMomentumFlux;
207	>	break;
208	>	case rnemdKePx:
209	>	case rnemdKePy:
210	>	hasCorrectFlux = hasMomentumFlux && hasKineticFlux;
211	>	break;
212	>	default:
213	>	methodFluxMismatch = true;
214	>	break;
215	>	}
216	>	break;
217	>	case rnemdVSS:
218	>	switch (rnemdFluxType_) {
219	>	case rnemdKE:
220	>	case rnemdRotKE:
221	>	case rnemdFullKE:
222	>	hasCorrectFlux = hasKineticFlux;
223	>	break;
224	>	case rnemdPx:
225	>	case rnemdPy:
226	>	case rnemdPz:
227	>	hasCorrectFlux = hasMomentumFlux;
228	>	break;
229	>	case rnemdLx:
230	>	case rnemdLy:
231	>	case rnemdLz:
232	>	hasCorrectFlux = hasAngularMomentumFlux;
233	>	break;
234	>	case rnemdPvector:
235	>	hasCorrectFlux = hasMomentumFluxVector;
236	>	break;
237	>	case rnemdLvector:
238	>	hasCorrectFlux = hasAngularMomentumFluxVector;
239	>	break;
240	>	case rnemdKePx:
241	>	case rnemdKePy:
242	>	hasCorrectFlux = hasMomentumFlux && hasKineticFlux;
243	>	break;
244	>	case rnemdKeLx:
245	>	case rnemdKeLy:
246	>	case rnemdKeLz:
247	>	hasCorrectFlux = hasAngularMomentumFlux && hasKineticFlux;
248	>	break;
249	>	case rnemdKePvector:
250	>	hasCorrectFlux = hasMomentumFluxVector && hasKineticFlux;
251	>	break;
252	>	case rnemdKeLvector:
253	>	hasCorrectFlux = hasAngularMomentumFluxVector && hasKineticFlux;
254	>	break;
255	>	default:
256	>	methodFluxMismatch = true;
257	>	break;
258	>	}
259	>	default:
260	>	break;
261		}
150	–	#endif
262
263	<	set_RNEMD_exchange_time(simParams->getRNEMD_exchangeTime());
264	<	set_RNEMD_nBins(simParams->getRNEMD_nBins());
265	<	midBin_ = nBins_ / 2;
266	<	if (simParams->haveRNEMD_logWidth()) {
267	<	rnemdLogWidth_ = simParams->getRNEMD_logWidth();
268	<	if (rnemdLogWidth_ != nBins_ && rnemdLogWidth_ != midBin_ + 1) {
269	<	std::cerr << "WARNING! RNEMD_logWidth has abnormal value!\n";
270	<	std::cerr << "Automaically set back to default.\n";
271	<	rnemdLogWidth_ = nBins_;
272	<	}
273	<	} else {
274	<	rnemdLogWidth_ = nBins_;
275	<	}
276	<	valueHist_.resize(rnemdLogWidth_, 0.0);
277	<	valueCount_.resize(rnemdLogWidth_, 0);
278	<	xTempHist_.resize(rnemdLogWidth_, 0.0);
279	<	yTempHist_.resize(rnemdLogWidth_, 0.0);
280	<	zTempHist_.resize(rnemdLogWidth_, 0.0);
263	>	if (methodFluxMismatch) {
264	>	sprintf(painCave.errMsg,
265	>	"RNEMD: The current method,\n"
266	>	"\t\t%s\n"
267	>	"\tcannot be used with the current flux type, %s\n",
268	>	methStr.c_str(), fluxStr.c_str());
269	>	painCave.isFatal = 1;
270	>	painCave.severity = OPENMD_ERROR;
271	>	simError();
272	>	}
273	>	if (!hasCorrectFlux) {
274	>	sprintf(painCave.errMsg,
275	>	"RNEMD: The current method, %s, and flux type, %s,\n"
276	>	"\tdid not have the correct flux value specified. Options\n"
277	>	"\tinclude: kineticFlux, momentumFlux, angularMomentumFlux,\n"
278	>	"\tmomentumFluxVector, and angularMomentumFluxVector.\n",
279	>	methStr.c_str(), fluxStr.c_str());
280	>	painCave.isFatal = 1;
281	>	painCave.severity = OPENMD_ERROR;
282	>	simError();
283	>	}
284
285	<	set_RNEMD_exchange_total(0.0);
286	<	if (simParams->haveRNEMD_targetFlux()) {
287	<	set_RNEMD_target_flux(simParams->getRNEMD_targetFlux());
285	>	if (hasKineticFlux) {
286	>	// convert the kcal / mol / Angstroms^2 / fs values in the md file
287	>	// into  amu / fs^3:
288	>	kineticFlux_ = rnemdParams->getKineticFlux()
289	>	* PhysicalConstants::energyConvert;
290		} else {
291	<	set_RNEMD_target_flux(0.0);
291	>	kineticFlux_ = 0.0;
292	>	}
293	>	if (hasMomentumFluxVector) {
294	>	momentumFluxVector_ = rnemdParams->getMomentumFluxVector();
295	>	} else {
296	>	momentumFluxVector_ = V3Zero;
297	>	if (hasMomentumFlux) {
298	>	RealType momentumFlux = rnemdParams->getMomentumFlux();
299	>	switch (rnemdFluxType_) {
300	>	case rnemdPx:
301	>	momentumFluxVector_.x() = momentumFlux;
302	>	break;
303	>	case rnemdPy:
304	>	momentumFluxVector_.y() = momentumFlux;
305	>	break;
306	>	case rnemdPz:
307	>	momentumFluxVector_.z() = momentumFlux;
308	>	break;
309	>	case rnemdKePx:
310	>	momentumFluxVector_.x() = momentumFlux;
311	>	break;
312	>	case rnemdKePy:
313	>	momentumFluxVector_.y() = momentumFlux;
314	>	break;
315	>	default:
316	>	break;
317	>	}
318	>	}
319	>	if (hasAngularMomentumFluxVector) {
320	>	angularMomentumFluxVector_ = rnemdParams->getAngularMomentumFluxVector();
321	>	} else {
322	>	angularMomentumFluxVector_ = V3Zero;
323	>	if (hasAngularMomentumFlux) {
324	>	RealType angularMomentumFlux = rnemdParams->getAngularMomentumFlux();
325	>	switch (rnemdFluxType_) {
326	>	case rnemdLx:
327	>	angularMomentumFluxVector_.x() = angularMomentumFlux;
328	>	break;
329	>	case rnemdLy:
330	>	angularMomentumFluxVector_.y() = angularMomentumFlux;
331	>	break;
332	>	case rnemdLz:
333	>	angularMomentumFluxVector_.z() = angularMomentumFlux;
334	>	break;
335	>	case rnemdKeLx:
336	>	angularMomentumFluxVector_.x() = angularMomentumFlux;
337	>	break;
338	>	case rnemdKeLy:
339	>	angularMomentumFluxVector_.y() = angularMomentumFlux;
340	>	break;
341	>	case rnemdKeLz:
342	>	angularMomentumFluxVector_.z() = angularMomentumFlux;
343	>	break;
344	>	default:
345	>	break;
346	>	}
347	>	}
348	>	}
349	>
350	>	if (hasCoordinateOrigin) {
351	>	coordinateOrigin_ = rnemdParams->getCoordinateOrigin();
352	>	} else {
353	>	coordinateOrigin_ = V3Zero;
354	>	}
355	>
356	>	// do some sanity checking
357	>
358	>	int selectionCount = seleMan_.getSelectionCount();
359	>
360	>	int nIntegrable = info->getNGlobalIntegrableObjects();
361	>
362	>	if (selectionCount > nIntegrable) {
363	>	sprintf(painCave.errMsg,
364	>	"RNEMD: The current objectSelection,\n"
365	>	"\t\t%s\n"
366	>	"\thas resulted in %d selected objects.  However,\n"
367	>	"\tthe total number of integrable objects in the system\n"
368	>	"\tis only %d.  This is almost certainly not what you want\n"
369	>	"\tto do.  A likely cause of this is forgetting the _RB_0\n"
370	>	"\tselector in the selection script!\n",
371	>	rnemdObjectSelection_.c_str(),
372	>	selectionCount, nIntegrable);
373	>	painCave.isFatal = 0;
374	>	painCave.severity = OPENMD_WARNING;
375	>	simError();
376	>	}
377	>
378	>	areaAccumulator_ = new Accumulator();
379	>
380	>	nBins_ = rnemdParams->getOutputBins();
381	>	binWidth_ = rnemdParams->getOutputBinWidth();
382	>
383	>	data_.resize(RNEMD::ENDINDEX);
384	>	OutputData z;
385	>	z.units =  "Angstroms";
386	>	z.title =  "Z";
387	>	z.dataType = "RealType";
388	>	z.accumulator.reserve(nBins_);
389	>	for (int i = 0; i < nBins_; i++)
390	>	z.accumulator.push_back( new Accumulator() );
391	>	data_[Z] = z;
392	>	outputMap_["Z"] =  Z;
393	>
394	>	OutputData r;
395	>	r.units =  "Angstroms";
396	>	r.title =  "R";
397	>	r.dataType = "RealType";
398	>	r.accumulator.reserve(nBins_);
399	>	for (int i = 0; i < nBins_; i++)
400	>	r.accumulator.push_back( new Accumulator() );
401	>	data_[R] = r;
402	>	outputMap_["R"] =  R;
403	>
404	>	OutputData temperature;
405	>	temperature.units =  "K";
406	>	temperature.title =  "Temperature";
407	>	temperature.dataType = "RealType";
408	>	temperature.accumulator.reserve(nBins_);
409	>	for (int i = 0; i < nBins_; i++)
410	>	temperature.accumulator.push_back( new Accumulator() );
411	>	data_[TEMPERATURE] = temperature;
412	>	outputMap_["TEMPERATURE"] =  TEMPERATURE;
413	>
414	>	OutputData velocity;
415	>	velocity.units = "angstroms/fs";
416	>	velocity.title =  "Velocity";
417	>	velocity.dataType = "Vector3d";
418	>	velocity.accumulator.reserve(nBins_);
419	>	for (int i = 0; i < nBins_; i++)
420	>	velocity.accumulator.push_back( new VectorAccumulator() );
421	>	data_[VELOCITY] = velocity;
422	>	outputMap_["VELOCITY"] = VELOCITY;
423	>
424	>	OutputData angularVelocity;
425	>	angularVelocity.units = "angstroms^2/fs";
426	>	angularVelocity.title =  "AngularVelocity";
427	>	angularVelocity.dataType = "Vector3d";
428	>	angularVelocity.accumulator.reserve(nBins_);
429	>	for (int i = 0; i < nBins_; i++)
430	>	angularVelocity.accumulator.push_back( new VectorAccumulator() );
431	>	data_[ANGULARVELOCITY] = angularVelocity;
432	>	outputMap_["ANGULARVELOCITY"] = ANGULARVELOCITY;
433	>
434	>	OutputData density;
435	>	density.units =  "g cm^-3";
436	>	density.title =  "Density";
437	>	density.dataType = "RealType";
438	>	density.accumulator.reserve(nBins_);
439	>	for (int i = 0; i < nBins_; i++)
440	>	density.accumulator.push_back( new Accumulator() );
441	>	data_[DENSITY] = density;
442	>	outputMap_["DENSITY"] =  DENSITY;
443	>
444	>	if (hasOutputFields) {
445	>	parseOutputFileFormat(rnemdParams->getOutputFields());
446	>	} else {
447	>	if (usePeriodicBoundaryConditions_)
448	>	outputMask_.set(Z);
449	>	else
450	>	outputMask_.set(R);
451	>	switch (rnemdFluxType_) {
452	>	case rnemdKE:
453	>	case rnemdRotKE:
454	>	case rnemdFullKE:
455	>	outputMask_.set(TEMPERATURE);
456	>	break;
457	>	case rnemdPx:
458	>	case rnemdPy:
459	>	outputMask_.set(VELOCITY);
460	>	break;
461	>	case rnemdPz:
462	>	case rnemdPvector:
463	>	outputMask_.set(VELOCITY);
464	>	outputMask_.set(DENSITY);
465	>	break;
466	>	case rnemdLx:
467	>	case rnemdLy:
468	>	case rnemdLz:
469	>	case rnemdLvector:
470	>	outputMask_.set(ANGULARVELOCITY);
471	>	break;
472	>	case rnemdKeLx:
473	>	case rnemdKeLy:
474	>	case rnemdKeLz:
475	>	case rnemdKeLvector:
476	>	outputMask_.set(TEMPERATURE);
477	>	outputMask_.set(ANGULARVELOCITY);
478	>	break;
479	>	case rnemdKePx:
480	>	case rnemdKePy:
481	>	outputMask_.set(TEMPERATURE);
482	>	outputMask_.set(VELOCITY);
483	>	break;
484	>	case rnemdKePvector:
485	>	outputMask_.set(TEMPERATURE);
486	>	outputMask_.set(VELOCITY);
487	>	outputMask_.set(DENSITY);
488	>	break;
489	>	default:
490	>	break;
491	>	}
492	>	}
493	>
494	>	if (hasOutputFileName) {
495	>	rnemdFileName_ = rnemdParams->getOutputFileName();
496	>	} else {
497	>	rnemdFileName_ = getPrefix(info->getFinalConfigFileName()) + ".rnemd";
498	>	}
499	>
500	>	exchangeTime_ = rnemdParams->getExchangeTime();
501	>
502	>	Snapshot* currentSnap_ = info->getSnapshotManager()->getCurrentSnapshot();
503	>	// total exchange sums are zeroed out at the beginning:
504	>
505	>	kineticExchange_ = 0.0;
506	>	momentumExchange_ = V3Zero;
507	>	angularMomentumExchange_ = V3Zero;
508	>
509	>	std::ostringstream selectionAstream;
510	>	std::ostringstream selectionBstream;
511	>
512	>	if (hasSelectionA_) {
513	>	selectionA_ = rnemdParams->getSelectionA();
514	>	} else {
515	>	if (usePeriodicBoundaryConditions_) {
516	>	Mat3x3d hmat = currentSnap_->getHmat();
517	>
518	>	if (hasSlabWidth)
519	>	slabWidth_ = rnemdParams->getSlabWidth();
520	>	else
521	>	slabWidth_ = hmat(2,2) / 10.0;
522	>
523	>	if (hasSlabACenter)
524	>	slabACenter_ = rnemdParams->getSlabACenter();
525	>	else
526	>	slabACenter_ = 0.0;
527	>
528	>	selectionAstream << "select wrappedz > "
529	>	<< slabACenter_ - 0.5*slabWidth_
530	>	<<  " && wrappedz < "
531	>	<< slabACenter_ + 0.5*slabWidth_;
532	>	selectionA_ = selectionAstream.str();
533	>	} else {
534	>	if (hasSphereARadius)
535	>	sphereARadius_ = rnemdParams->getSphereARadius();
536	>	else {
537	>	// use an initial guess to the size of the inner slab to be 1/10 the
538	>	// radius of an approximately spherical hull:
539	>	Thermo thermo(info);
540	>	RealType hVol = thermo.getHullVolume();
541	>	sphereARadius_ = 0.1 * pow((3.0 * hVol / (4.0 * M_PI)), 1.0/3.0);
542	>	}
543	>	selectionAstream << "select r < " << sphereARadius_;
544	>	selectionA_ = selectionAstream.str();
545	>	}
546	>	}
547	>
548	>	if (hasSelectionB_) {
549	>	selectionB_ = rnemdParams->getSelectionB();
550	>
551	>	} else {
552	>	if (usePeriodicBoundaryConditions_) {
553	>	Mat3x3d hmat = currentSnap_->getHmat();
554	>
555	>	if (hasSlabWidth)
556	>	slabWidth_ = rnemdParams->getSlabWidth();
557	>	else
558	>	slabWidth_ = hmat(2,2) / 10.0;
559	>
560	>	if (hasSlabBCenter)
561	>	slabBCenter_ = rnemdParams->getSlabBCenter();
562	>	else
563	>	slabBCenter_ = hmat(2,2) / 2.0;
564	>
565	>	selectionBstream << "select wrappedz > "
566	>	<< slabBCenter_ - 0.5*slabWidth_
567	>	<<  " && wrappedz < "
568	>	<< slabBCenter_ + 0.5*slabWidth_;
569	>	selectionB_ = selectionBstream.str();
570	>	} else {
571	>	if (hasSphereBRadius_) {
572	>	sphereBRadius_ = rnemdParams->getSphereBRadius();
573	>	selectionBstream << "select r > " << sphereBRadius_;
574	>	selectionB_ = selectionBstream.str();
575	>	} else {
576	>	selectionB_ = "select hull";
577	>	BisHull_ = true;
578	>	hasSelectionB_ = true;
579	>	}
580	>	}
581	>	}
582		}
583
584	<	#ifndef IS_MPI
585	<	if (simParams->haveSeed()) {
586	<	seedValue = simParams->getSeed();
587	<	randNumGen_ = new SeqRandNumGen(seedValue);
588	<	}else {
589	<	randNumGen_ = new SeqRandNumGen();
590	<	}
591	<	#else
592	<	if (simParams->haveSeed()) {
187	<	seedValue = simParams->getSeed();
188	<	randNumGen_ = new ParallelRandNumGen(seedValue);
189	<	}else {
190	<	randNumGen_ = new ParallelRandNumGen();
191	<	}
192	<	#endif
584	>	// object evaluator:
585	>	evaluator_.loadScriptString(rnemdObjectSelection_);
586	>	seleMan_.setSelectionSet(evaluator_.evaluate());
587	>	evaluatorA_.loadScriptString(selectionA_);
588	>	evaluatorB_.loadScriptString(selectionB_);
589	>	seleManA_.setSelectionSet(evaluatorA_.evaluate());
590	>	seleManB_.setSelectionSet(evaluatorB_.evaluate());
591	>	commonA_ = seleManA_ & seleMan_;
592	>	commonB_ = seleManB_ & seleMan_;
593		}
594
195	–	RNEMD::~RNEMD() {
196	–	delete randNumGen_;
595
596	+	RNEMD::~RNEMD() {
597	+	if (!doRNEMD_) return;
598		#ifdef IS_MPI
599		if (worldRank == 0) {
600		#endif
601	<	std::cerr << "total fail trials: " << failTrialCount_ << "\n";
602	<	rnemdLog_.close();
603	<	if (rnemdType_ == rnemdKineticScale || rnemdType_ == rnemdPxScale || rnemdType_ == rnemdPyScale)
604	<	std::cerr<< "total root-checking warnings: " << failRootCount_ << "\n";
605	<	if (rnemdType_ == rnemdPx || rnemdType_ == rnemdPxScale || rnemdType_ == rnemdPy || rnemdType_ == rnemdPyScale) {
206	<	xTempLog_.close();
207	<	yTempLog_.close();
208	<	zTempLog_.close();
209	<	}
601	>
602	>	writeOutputFile();
603	>
604	>	rnemdFile_.close();
605	>
606		#ifdef IS_MPI
607		}
608		#endif
609	+
610	+	// delete all of the objects we created:
611	+	delete areaAccumulator_;
612	+	data_.clear();
613		}
614	+
615	+	void RNEMD::doSwap(SelectionManager& smanA, SelectionManager& smanB) {
616	+	if (!doRNEMD_) return;
617	+	int selei;
618	+	int selej;
619
215	–	void RNEMD::doSwap() {
216	–
620		Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
621		Mat3x3d hmat = currentSnap_->getHmat();
622
220	–	seleMan_.setSelectionSet(evaluator_.evaluate());
221	–
222	–	int selei;
623		StuntDouble* sd;
224	–	int idx;
624
625		RealType min_val;
626	<	bool min_found = false;
626	>	int min_found = 0;
627		StuntDouble* min_sd;
628
629		RealType max_val;
630	<	bool max_found = false;
630	>	int max_found = 0;
631		StuntDouble* max_sd;
632
633	<	for (sd = seleMan_.beginSelected(selei); sd != NULL;
634	<	sd = seleMan_.nextSelected(selei)) {
633	>	for (sd = seleManA_.beginSelected(selei); sd != NULL;
634	>	sd = seleManA_.nextSelected(selei)) {
635
237	–	idx = sd->getLocalIndex();
238	–
636		Vector3d pos = sd->getPos();
637	<
637	>
638		// wrap the stuntdouble's position back into the box:
639	<
639	>
640		if (usePeriodicBoundaryConditions_)
641		currentSnap_->wrapVector(pos);
642	<
643	<	// which bin is this stuntdouble in?
644	<	// wrapped positions are in the range [-0.5*hmat(2,2), +0.5*hmat(2,2)]
645	<
646	<	int binNo = int(nBins_ * (pos.z() / hmat(2,2) + 0.5)) % nBins_;
647	<
648	<
252	<	// if we're in bin 0 or the middleBin
253	<	if (binNo == 0 || binNo == midBin_) {
642	>
643	>	RealType mass = sd->getMass();
644	>	Vector3d vel = sd->getVel();
645	>	RealType value;
646	>
647	>	switch(rnemdFluxType_) {
648	>	case rnemdKE :
649
650	<	RealType mass = sd->getMass();
651	<	Vector3d vel = sd->getVel();
652	<	RealType value;
653	<
654	<	switch(rnemdType_) {
260	<	case rnemdKineticSwap :
650	>	value = mass * vel.lengthSquare();
651	>
652	>	if (sd->isDirectional()) {
653	>	Vector3d angMom = sd->getJ();
654	>	Mat3x3d I = sd->getI();
655
656	<	value = mass * (vel[0]*vel[0] + vel[1]*vel[1] +
657	<	vel[2]*vel[2]);
658	<	if (sd->isDirectional()) {
659	<	Vector3d angMom = sd->getJ();
660	<	Mat3x3d I = sd->getI();
661	<
662	<	if (sd->isLinear()) {
663	<	int i = sd->linearAxis();
664	<	int j = (i + 1) % 3;
665	<	int k = (i + 2) % 3;
272	<	value += angMom[j] * angMom[j] / I(j, j) +
273	<	angMom[k] * angMom[k] / I(k, k);
274	<	} else {
275	<	value += angMom[0]*angMom[0]/I(0, 0)
276	<	+ angMom[1]*angMom[1]/I(1, 1)
277	<	+ angMom[2]*angMom[2]/I(2, 2);
278	<	}
656	>	if (sd->isLinear()) {
657	>	int i = sd->linearAxis();
658	>	int j = (i + 1) % 3;
659	>	int k = (i + 2) % 3;
660	>	value += angMom[j] * angMom[j] / I(j, j) +
661	>	angMom[k] * angMom[k] / I(k, k);
662	>	} else {
663	>	value += angMom[0]*angMom[0]/I(0, 0)
664	>	+ angMom[1]*angMom[1]/I(1, 1)
665	>	+ angMom[2]*angMom[2]/I(2, 2);
666		}
667	<	//make exchangeSum_ comparable between swap & scale
668	<	//temporarily without using energyConvert
669	<	//value = value * 0.5 / PhysicalConstants::energyConvert;
670	<	value *= 0.5;
671	<	break;
672	<	case rnemdPx :
673	<	value = mass * vel[0];
674	<	break;
675	<	case rnemdPy :
676	<	value = mass * vel[1];
677	<	break;
678	<	case rnemdPz :
679	<	value = mass * vel[2];
680	<	break;
681	<	default :
682	<	break;
667	>	} //angular momenta exchange enabled
668	>	value *= 0.5;
669	>	break;
670	>	case rnemdPx :
671	>	value = mass * vel[0];
672	>	break;
673	>	case rnemdPy :
674	>	value = mass * vel[1];
675	>	break;
676	>	case rnemdPz :
677	>	value = mass * vel[2];
678	>	break;
679	>	default :
680	>	break;
681	>	}
682	>	if (!max_found) {
683	>	max_val = value;
684	>	max_sd = sd;
685	>	max_found = 1;
686	>	} else {
687	>	if (max_val < value) {
688	>	max_val = value;
689	>	max_sd = sd;
690		}
691	+	}
692	+	}
693
694	<	if (binNo == 0) {
695	<	if (!min_found) {
696	<	min_val = value;
697	<	min_sd = sd;
698	<	min_found = true;
699	<	} else {
700	<	if (min_val > value) {
701	<	min_val = value;
702	<	min_sd = sd;
703	<	}
704	<	}
705	<	} else { //midBin_
706	<	if (!max_found) {
707	<	max_val = value;
708	<	max_sd = sd;
709	<	max_found = true;
710	<	} else {
711	<	if (max_val < value) {
712	<	max_val = value;
713	<	max_sd = sd;
714	<	}
715	<	}
716	<	}
694	>	for (sd = seleManB_.beginSelected(selej); sd != NULL;
695	>	sd = seleManB_.nextSelected(selej)) {
696	>
697	>	Vector3d pos = sd->getPos();
698	>
699	>	// wrap the stuntdouble's position back into the box:
700	>
701	>	if (usePeriodicBoundaryConditions_)
702	>	currentSnap_->wrapVector(pos);
703	>
704	>	RealType mass = sd->getMass();
705	>	Vector3d vel = sd->getVel();
706	>	RealType value;
707	>
708	>	switch(rnemdFluxType_) {
709	>	case rnemdKE :
710	>
711	>	value = mass * vel.lengthSquare();
712	>
713	>	if (sd->isDirectional()) {
714	>	Vector3d angMom = sd->getJ();
715	>	Mat3x3d I = sd->getI();
716	>
717	>	if (sd->isLinear()) {
718	>	int i = sd->linearAxis();
719	>	int j = (i + 1) % 3;
720	>	int k = (i + 2) % 3;
721	>	value += angMom[j] * angMom[j] / I(j, j) +
722	>	angMom[k] * angMom[k] / I(k, k);
723	>	} else {
724	>	value += angMom[0]*angMom[0]/I(0, 0)
725	>	+ angMom[1]*angMom[1]/I(1, 1)
726	>	+ angMom[2]*angMom[2]/I(2, 2);
727	>	}
728	>	} //angular momenta exchange enabled
729	>	value *= 0.5;
730	>	break;
731	>	case rnemdPx :
732	>	value = mass * vel[0];
733	>	break;
734	>	case rnemdPy :
735	>	value = mass * vel[1];
736	>	break;
737	>	case rnemdPz :
738	>	value = mass * vel[2];
739	>	break;
740	>	default :
741	>	break;
742		}
743	+
744	+	if (!min_found) {
745	+	min_val = value;
746	+	min_sd = sd;
747	+	min_found = 1;
748	+	} else {
749	+	if (min_val > value) {
750	+	min_val = value;
751	+	min_sd = sd;
752	+	}
753	+	}
754		}
755	+
756	+	#ifdef IS_MPI
757	+	int worldRank;
758	+	MPI_Comm_rank( MPI_COMM_WORLD, &worldRank);
759	+
760	+	int my_min_found = min_found;
761	+	int my_max_found = max_found;
762
324	–	#ifdef IS_MPI
325	–	int nProc, worldRank;
326	–
327	–	nProc = MPI::COMM_WORLD.Get_size();
328	–	worldRank = MPI::COMM_WORLD.Get_rank();
329	–
330	–	bool my_min_found = min_found;
331	–	bool my_max_found = max_found;
332	–
763		// Even if we didn't find a minimum, did someone else?
764	<	MPI::COMM_WORLD.Allreduce(&my_min_found, &min_found,
765	<	1, MPI::BOOL, MPI::LAND);
336	<
764	>	MPI_Allreduce(&my_min_found, &min_found, 1, MPI_INT, MPI_LOR,
765	>	MPI_COMM_WORLD);
766		// Even if we didn't find a maximum, did someone else?
767	<	MPI::COMM_WORLD.Allreduce(&my_max_found, &max_found,
768	<	1, MPI::BOOL, MPI::LAND);
769	<
770	<	struct {
771	<	RealType val;
772	<	int rank;
773	<	} max_vals, min_vals;
774	<
775	<	if (min_found) {
776	<	if (my_min_found)
767	>	MPI_Allreduce(&my_max_found, &max_found, 1, MPI_INT, MPI_LOR,
768	>	MPI_COMM_WORLD);
769	>	#endif
770	>
771	>	if (max_found && min_found) {
772	>
773	>	#ifdef IS_MPI
774	>	struct {
775	>	RealType val;
776	>	int rank;
777	>	} max_vals, min_vals;
778	>
779	>	if (my_min_found) {
780		min_vals.val = min_val;
781	<	else
781	>	} else {
782		min_vals.val = HONKING_LARGE_VALUE;
783	<
783	>	}
784		min_vals.rank = worldRank;
785
786		// Who had the minimum?
787	<	MPI::COMM_WORLD.Allreduce(&min_vals, &min_vals,
788	<	1, MPI::REALTYPE_INT, MPI::MINLOC);
787	>	MPI_Allreduce(&min_vals, &min_vals,
788	>	1, MPI_REALTYPE_INT, MPI_MINLOC, MPI_COMM_WORLD);
789		min_val = min_vals.val;
358	–	}
790
791	<	if (max_found) {
361	<	if (my_max_found)
791	>	if (my_max_found) {
792		max_vals.val = max_val;
793	<	else
793	>	} else {
794		max_vals.val = -HONKING_LARGE_VALUE;
795	<
795	>	}
796		max_vals.rank = worldRank;
797
798		// Who had the maximum?
799	<	MPI::COMM_WORLD.Allreduce(&max_vals, &max_vals,
800	<	1, MPI::REALTYPE_INT, MPI::MAXLOC);
799	>	MPI_Allreduce(&max_vals, &max_vals,
800	>	1, MPI_REALTYPE_INT, MPI_MAXLOC, MPI_COMM_WORLD);
801		max_val = max_vals.val;
372	–	}
802		#endif
803	<
804	<	if (max_found && min_found) {
805	<	if (min_val< max_val) {
377	<
803	>
804	>	if (min_val < max_val) {
805	>
806		#ifdef IS_MPI
807		if (max_vals.rank == worldRank && min_vals.rank == worldRank) {
808		// I have both maximum and minimum, so proceed like a single
809		// processor version:
810		#endif
811	<	// objects to be swapped: velocity & angular velocity
811	>
812		Vector3d min_vel = min_sd->getVel();
813		Vector3d max_vel = max_sd->getVel();
814		RealType temp_vel;
815
816	<	switch(rnemdType_) {
817	<	case rnemdKineticSwap :
816	>	switch(rnemdFluxType_) {
817	>	case rnemdKE :
818		min_sd->setVel(max_vel);
819		max_sd->setVel(min_vel);
820	<	if (min_sd->isDirectional() && max_sd->isDirectional()) {
820	>	if (min_sd->isDirectional() && max_sd->isDirectional()) {
821		Vector3d min_angMom = min_sd->getJ();
822		Vector3d max_angMom = max_sd->getJ();
823		min_sd->setJ(max_angMom);
824		max_sd->setJ(min_angMom);
825	<	}
825	>	}//angular momenta exchange enabled
826	>	//assumes same rigid body identity
827		break;
828		case rnemdPx :
829		temp_vel = min_vel.x();
#	Line 420 | Line 849 | namespace OpenMD {
849		default :
850		break;
851		}
852	+
853		#ifdef IS_MPI
854		// the rest of the cases only apply in parallel simulations:
855		} else if (max_vals.rank == worldRank) {
#	Line 427 | Line 857 | namespace OpenMD {
857
858		Vector3d min_vel;
859		Vector3d max_vel = max_sd->getVel();
860	<	MPI::Status status;
860	>	MPI_Status* status;
861
862		// point-to-point swap of the velocity vector
863	<	MPI::COMM_WORLD.Sendrecv(max_vel.getArrayPointer(), 3, MPI::REALTYPE,
864	<	min_vals.rank, 0,
865	<	min_vel.getArrayPointer(), 3, MPI::REALTYPE,
866	<	min_vals.rank, 0, status);
863	>	MPI_Sendrecv(max_vel.getArrayPointer(), 3, MPI_REALTYPE,
864	>	min_vals.rank, 0,
865	>	min_vel.getArrayPointer(), 3, MPI_REALTYPE,
866	>	min_vals.rank, 0, MPI_COMM_WORLD, status);
867
868	<	switch(rnemdType_) {
869	<	case rnemdKineticSwap :
868	>	switch(rnemdFluxType_) {
869	>	case rnemdKE :
870		max_sd->setVel(min_vel);
871	<
871	>	//angular momenta exchange enabled
872		if (max_sd->isDirectional()) {
873		Vector3d min_angMom;
874		Vector3d max_angMom = max_sd->getJ();
875	<
875	>
876		// point-to-point swap of the angular momentum vector
877	<	MPI::COMM_WORLD.Sendrecv(max_angMom.getArrayPointer(), 3,
878	<	MPI::REALTYPE, min_vals.rank, 1,
879	<	min_angMom.getArrayPointer(), 3,
880	<	MPI::REALTYPE, min_vals.rank, 1,
881	<	status);
882	<
877	>	MPI_Sendrecv(max_angMom.getArrayPointer(), 3,
878	>	MPI_REALTYPE, min_vals.rank, 1,
879	>	min_angMom.getArrayPointer(), 3,
880	>	MPI_REALTYPE, min_vals.rank, 1,
881	>	MPI_COMM_WORLD, status);
882	>
883		max_sd->setJ(min_angMom);
884	<	}
884	>	}
885		break;
886		case rnemdPx :
887		max_vel.x() = min_vel.x();
#	Line 473 | Line 903 | namespace OpenMD {
903
904		Vector3d max_vel;
905		Vector3d min_vel = min_sd->getVel();
906	<	MPI::Status status;
906	>	MPI_Status* status;
907
908		// point-to-point swap of the velocity vector
909	<	MPI::COMM_WORLD.Sendrecv(min_vel.getArrayPointer(), 3, MPI::REALTYPE,
910	<	max_vals.rank, 0,
911	<	max_vel.getArrayPointer(), 3, MPI::REALTYPE,
912	<	max_vals.rank, 0, status);
909	>	MPI_Sendrecv(min_vel.getArrayPointer(), 3, MPI_REALTYPE,
910	>	max_vals.rank, 0,
911	>	max_vel.getArrayPointer(), 3, MPI_REALTYPE,
912	>	max_vals.rank, 0, MPI_COMM_WORLD, status);
913
914	<	switch(rnemdType_) {
915	<	case rnemdKineticSwap :
914	>	switch(rnemdFluxType_) {
915	>	case rnemdKE :
916		min_sd->setVel(max_vel);
917	<
917	>	//angular momenta exchange enabled
918		if (min_sd->isDirectional()) {
919		Vector3d min_angMom = min_sd->getJ();
920		Vector3d max_angMom;
921	<
921	>
922		// point-to-point swap of the angular momentum vector
923	<	MPI::COMM_WORLD.Sendrecv(min_angMom.getArrayPointer(), 3,
924	<	MPI::REALTYPE, max_vals.rank, 1,
925	<	max_angMom.getArrayPointer(), 3,
926	<	MPI::REALTYPE, max_vals.rank, 1,
927	<	status);
928	<
923	>	MPI_Sendrecv(min_angMom.getArrayPointer(), 3,
924	>	MPI_REALTYPE, max_vals.rank, 1,
925	>	max_angMom.getArrayPointer(), 3,
926	>	MPI_REALTYPE, max_vals.rank, 1,
927	>	MPI_COMM_WORLD, status);
928	>
929		min_sd->setJ(max_angMom);
930		}
931		break;
#	Line 516 | Line 946 | namespace OpenMD {
946		}
947		}
948		#endif
949	<	exchangeSum_ += max_val - min_val;
950	<	} else {
951	<	std::cerr << "exchange NOT performed!\nmin_val > max_val.\n";
949	>
950	>	switch(rnemdFluxType_) {
951	>	case rnemdKE:
952	>	kineticExchange_ += max_val - min_val;
953	>	break;
954	>	case rnemdPx:
955	>	momentumExchange_.x() += max_val - min_val;
956	>	break;
957	>	case rnemdPy:
958	>	momentumExchange_.y() += max_val - min_val;
959	>	break;
960	>	case rnemdPz:
961	>	momentumExchange_.z() += max_val - min_val;
962	>	break;
963	>	default:
964	>	break;
965	>	}
966	>	} else {
967	>	sprintf(painCave.errMsg,
968	>	"RNEMD::doSwap exchange NOT performed because min_val > max_val\n");
969	>	painCave.isFatal = 0;
970	>	painCave.severity = OPENMD_INFO;
971	>	simError();
972		failTrialCount_++;
973		}
974		} else {
975	<	std::cerr << "exchange NOT performed!\n";
976	<	std::cerr << "at least one of the two slabs empty.\n";
975	>	sprintf(painCave.errMsg,
976	>	"RNEMD::doSwap exchange NOT performed because selected object\n"
977	>	"\twas not present in at least one of the two slabs.\n");
978	>	painCave.isFatal = 0;
979	>	painCave.severity = OPENMD_INFO;
980	>	simError();
981		failTrialCount_++;
982	<	}
529	<
982	>	}
983		}
984
985	<	void RNEMD::doScale() {
985	>	void RNEMD::doNIVS(SelectionManager& smanA, SelectionManager& smanB) {
986	>	if (!doRNEMD_) return;
987	>	int selei;
988	>	int selej;
989
990		Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
991	+	RealType time = currentSnap_->getTime();
992		Mat3x3d hmat = currentSnap_->getHmat();
993
537	–	seleMan_.setSelectionSet(evaluator_.evaluate());
538	–
539	–	int selei;
994		StuntDouble* sd;
541	–	int idx;
995
996	<	std::vector<StuntDouble*> hotBin, coldBin;
996	>	vector<StuntDouble*> hotBin, coldBin;
997
998		RealType Phx = 0.0;
999		RealType Phy = 0.0;
#	Line 548 | Line 1001 | namespace OpenMD {
1001		RealType Khx = 0.0;
1002		RealType Khy = 0.0;
1003		RealType Khz = 0.0;
1004	+	RealType Khw = 0.0;
1005		RealType Pcx = 0.0;
1006		RealType Pcy = 0.0;
1007		RealType Pcz = 0.0;
1008		RealType Kcx = 0.0;
1009		RealType Kcy = 0.0;
1010		RealType Kcz = 0.0;
1011	+	RealType Kcw = 0.0;
1012
1013	<	for (sd = seleMan_.beginSelected(selei); sd != NULL;
1014	<	sd = seleMan_.nextSelected(selei)) {
1013	>	for (sd = smanA.beginSelected(selei); sd != NULL;
1014	>	sd = smanA.nextSelected(selei)) {
1015
561	–	idx = sd->getLocalIndex();
562	–
1016		Vector3d pos = sd->getPos();
1017	<
1017	>
1018		// wrap the stuntdouble's position back into the box:
1019	<
1019	>
1020		if (usePeriodicBoundaryConditions_)
1021		currentSnap_->wrapVector(pos);
1022	<
1023	<	// which bin is this stuntdouble in?
1024	<	// wrapped positions are in the range [-0.5*hmat(2,2), +0.5*hmat(2,2)]
1025	<
1026	<	int binNo = int(nBins_ * (pos.z() / hmat(2,2) + 0.5)) % nBins_;
1027	<
1028	<	// if we're in bin 0 or the middleBin
1029	<	if (binNo == 0 || binNo == midBin_) {
1030	<
1031	<	RealType mass = sd->getMass();
1032	<	Vector3d vel = sd->getVel();
1033	<
1034	<	if (binNo == 0) {
1035	<	hotBin.push_back(sd);
1036	<	Phx += mass * vel.x();
1037	<	Phy += mass * vel.y();
1038	<	Phz += mass * vel.z();
1039	<	Khx += mass * vel.x() * vel.x();
1040	<	Khy += mass * vel.y() * vel.y();
1041	<	Khz += mass * vel.z() * vel.z();
1042	<	} else { //midBin_
1043	<	coldBin.push_back(sd);
1044	<	Pcx += mass * vel.x();
1045	<	Pcy += mass * vel.y();
1046	<	Pcz += mass * vel.z();
1047	<	Kcx += mass * vel.x() * vel.x();
595	<	Kcy += mass * vel.y() * vel.y();
596	<	Kcz += mass * vel.z() * vel.z();
597	<	}
1022	>
1023	>
1024	>	RealType mass = sd->getMass();
1025	>	Vector3d vel = sd->getVel();
1026	>
1027	>	hotBin.push_back(sd);
1028	>	Phx += mass * vel.x();
1029	>	Phy += mass * vel.y();
1030	>	Phz += mass * vel.z();
1031	>	Khx += mass * vel.x() * vel.x();
1032	>	Khy += mass * vel.y() * vel.y();
1033	>	Khz += mass * vel.z() * vel.z();
1034	>	if (sd->isDirectional()) {
1035	>	Vector3d angMom = sd->getJ();
1036	>	Mat3x3d I = sd->getI();
1037	>	if (sd->isLinear()) {
1038	>	int i = sd->linearAxis();
1039	>	int j = (i + 1) % 3;
1040	>	int k = (i + 2) % 3;
1041	>	Khw += angMom[j] * angMom[j] / I(j, j) +
1042	>	angMom[k] * angMom[k] / I(k, k);
1043	>	} else {
1044	>	Khw += angMom[0]*angMom[0]/I(0, 0)
1045	>	+ angMom[1]*angMom[1]/I(1, 1)
1046	>	+ angMom[2]*angMom[2]/I(2, 2);
1047	>	}
1048		}
1049		}
1050	+	for (sd = smanB.beginSelected(selej); sd != NULL;
1051	+	sd = smanB.nextSelected(selej)) {
1052	+	Vector3d pos = sd->getPos();
1053	+
1054	+	// wrap the stuntdouble's position back into the box:
1055	+
1056	+	if (usePeriodicBoundaryConditions_)
1057	+	currentSnap_->wrapVector(pos);
1058	+
1059	+	RealType mass = sd->getMass();
1060	+	Vector3d vel = sd->getVel();
1061
1062	+	coldBin.push_back(sd);
1063	+	Pcx += mass * vel.x();
1064	+	Pcy += mass * vel.y();
1065	+	Pcz += mass * vel.z();
1066	+	Kcx += mass * vel.x() * vel.x();
1067	+	Kcy += mass * vel.y() * vel.y();
1068	+	Kcz += mass * vel.z() * vel.z();
1069	+	if (sd->isDirectional()) {
1070	+	Vector3d angMom = sd->getJ();
1071	+	Mat3x3d I = sd->getI();
1072	+	if (sd->isLinear()) {
1073	+	int i = sd->linearAxis();
1074	+	int j = (i + 1) % 3;
1075	+	int k = (i + 2) % 3;
1076	+	Kcw += angMom[j] * angMom[j] / I(j, j) +
1077	+	angMom[k] * angMom[k] / I(k, k);
1078	+	} else {
1079	+	Kcw += angMom[0]*angMom[0]/I(0, 0)
1080	+	+ angMom[1]*angMom[1]/I(1, 1)
1081	+	+ angMom[2]*angMom[2]/I(2, 2);
1082	+	}
1083	+	}
1084	+	}
1085	+
1086		Khx *= 0.5;
1087		Khy *= 0.5;
1088		Khz *= 0.5;
1089	+	Khw *= 0.5;
1090		Kcx *= 0.5;
1091		Kcy *= 0.5;
1092		Kcz *= 0.5;
1093	+	Kcw *= 0.5;
1094
1095		#ifdef IS_MPI
1096	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Phx, 1, MPI::REALTYPE, MPI::SUM);
1097	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Phy, 1, MPI::REALTYPE, MPI::SUM);
1098	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Phz, 1, MPI::REALTYPE, MPI::SUM);
1099	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Pcx, 1, MPI::REALTYPE, MPI::SUM);
1100	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Pcy, 1, MPI::REALTYPE, MPI::SUM);
1101	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Pcz, 1, MPI::REALTYPE, MPI::SUM);
1096	>	MPI_Allreduce(MPI_IN_PLACE, &Phx, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1097	>	MPI_Allreduce(MPI_IN_PLACE, &Phy, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1098	>	MPI_Allreduce(MPI_IN_PLACE, &Phz, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1099	>	MPI_Allreduce(MPI_IN_PLACE, &Pcx, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1100	>	MPI_Allreduce(MPI_IN_PLACE, &Pcy, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1101	>	MPI_Allreduce(MPI_IN_PLACE, &Pcz, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1102
1103	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Khx, 1, MPI::REALTYPE, MPI::SUM);
1104	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Khy, 1, MPI::REALTYPE, MPI::SUM);
1105	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Khz, 1, MPI::REALTYPE, MPI::SUM);
1106	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Kcx, 1, MPI::REALTYPE, MPI::SUM);
1107	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Kcy, 1, MPI::REALTYPE, MPI::SUM);
1108	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &Kcz, 1, MPI::REALTYPE, MPI::SUM);
1103	>	MPI_Allreduce(MPI_IN_PLACE, &Khx, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1104	>	MPI_Allreduce(MPI_IN_PLACE, &Khy, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1105	>	MPI_Allreduce(MPI_IN_PLACE, &Khz, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1106	>	MPI_Allreduce(MPI_IN_PLACE, &Khw, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1107	>
1108	>	MPI_Allreduce(MPI_IN_PLACE, &Kcx, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1109	>	MPI_Allreduce(MPI_IN_PLACE, &Kcy, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1110	>	MPI_Allreduce(MPI_IN_PLACE, &Kcz, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1111	>	MPI_Allreduce(MPI_IN_PLACE, &Kcw, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1112		#endif
1113
1114	<	//use coldBin coeff's
1114	>	//solve coldBin coeff's first
1115		RealType px = Pcx / Phx;
1116		RealType py = Pcy / Phy;
1117		RealType pz = Pcz / Phz;
1118	+	RealType c, x, y, z;
1119	+	bool successfulScale = false;
1120	+	if ((rnemdFluxType_ == rnemdFullKE) ||
1121	+	(rnemdFluxType_ == rnemdRotKE)) {
1122	+	//may need sanity check Khw & Kcw > 0
1123
1124	<	RealType a000, a110, c0, a001, a111, b01, b11, c1, c;
1125	<	switch(rnemdType_) {
1126	<	case rnemdKineticScale :
1127	<	/*used hotBin coeff's & only scale x & y dimensions
1128	<	RealType px = Phx / Pcx;
634	<	RealType py = Phy / Pcy;
635	<	a110 = Khy;
636	<	c0 = - Khx - Khy - targetFlux_;
637	<	a000 = Khx;
638	<	a111 = Kcy * py * py
639	<	b11 = -2.0 * Kcy * py * (1.0 + py);
640	<	c1 = Kcy * py * (2.0 + py) + Kcx * px * ( 2.0 + px) + targetFlux_;
641	<	b01 = -2.0 * Kcx * px * (1.0 + px);
642	<	a001 = Kcx * px * px;
643	<	*/
1124	>	if (rnemdFluxType_ == rnemdFullKE) {
1125	>	c = 1.0 - kineticTarget_ / (Kcx + Kcy + Kcz + Kcw);
1126	>	} else {
1127	>	c = 1.0 - kineticTarget_ / Kcw;
1128	>	}
1129
1130	<	//scale all three dimensions, let c_x = c_y
1131	<	a000 = Kcx + Kcy;
1132	<	a110 = Kcz;
1133	<	c0 = targetFlux_ - Kcx - Kcy - Kcz;
1134	<	a001 = Khx * px * px + Khy * py * py;
1135	<	a111 = Khz * pz * pz;
1136	<	b01 = -2.0 * (Khx * px * (1.0 + px) + Khy * py * (1.0 + py));
1137	<	b11 = -2.0 * Khz * pz * (1.0 + pz);
1138	<	c1 = Khx * px * (2.0 + px) + Khy * py * (2.0 + py)
1139	<	+ Khz * pz * (2.0 + pz) - targetFlux_;
1140	<	break;
1141	<	case rnemdPxScale :
1142	<	c = 1 - targetFlux_ / Pcx;
1143	<	a000 = Kcy;
1144	<	a110 = Kcz;
1145	<	c0 = Kcx * c * c - Kcx - Kcy - Kcz;
1146	<	a001 = py * py * Khy;
1147	<	a111 = pz * pz * Khz;
1148	<	b01 = -2.0 * Khy * py * (1.0 + py);
1149	<	b11 = -2.0 * Khz * pz * (1.0 + pz);
1150	<	c1 = Khy * py * (2.0 + py) + Khz * pz * (2.0 + pz)
1151	<	+ Khx * (fastpow(c * px - px - 1.0, 2) - 1.0);
1152	<	break;
1153	<	case rnemdPyScale :
1154	<	c = 1 - targetFlux_ / Pcy;
1155	<	a000 = Kcx;
1156	<	a110 = Kcz;
1157	<	c0 = Kcy * c * c - Kcx - Kcy - Kcz;
1158	<	a001 = px * px * Khx;
1159	<	a111 = pz * pz * Khz;
1160	<	b01 = -2.0 * Khx * px * (1.0 + px);
1161	<	b11 = -2.0 * Khz * pz * (1.0 + pz);
1162	<	c1 = Khx * px * (2.0 + px) + Khz * pz * (2.0 + pz)
1163	<	+ Khy * (fastpow(c * py - py - 1.0, 2) - 1.0);
1130	>	if ((c > 0.81) && (c < 1.21)) {//restrict scaling coefficients
1131	>	c = sqrt(c);
1132	>
1133	>	RealType w = 0.0;
1134	>	if (rnemdFluxType_ ==  rnemdFullKE) {
1135	>	x = 1.0 + px * (1.0 - c);
1136	>	y = 1.0 + py * (1.0 - c);
1137	>	z = 1.0 + pz * (1.0 - c);
1138	>	/* more complicated way
1139	>	w = 1.0 + (Kcw - Kcw * c * c - (c * c * (Kcx + Kcy + Kcz
1140	>	+ Khx * px * px + Khy * py * py + Khz * pz * pz)
1141	>	- 2.0 * c * (Khx * px * (1.0 + px) + Khy * py * (1.0 + py)
1142	>	+ Khz * pz * (1.0 + pz)) + Khx * px * (2.0 + px)
1143	>	+ Khy * py * (2.0 + py) + Khz * pz * (2.0 + pz)
1144	>	- Kcx - Kcy - Kcz)) / Khw; the following is simpler
1145	>	*/
1146	>	if ((fabs(x - 1.0) < 0.1) && (fabs(y - 1.0) < 0.1) &&
1147	>	(fabs(z - 1.0) < 0.1)) {
1148	>	w = 1.0 + (kineticTarget_
1149	>	+ Khx * (1.0 - x * x) + Khy * (1.0 - y * y)
1150	>	+ Khz * (1.0 - z * z)) / Khw;
1151	>	}//no need to calculate w if x, y or z is out of range
1152	>	} else {
1153	>	w = 1.0 + kineticTarget_ / Khw;
1154	>	}
1155	>	if ((w > 0.81) && (w < 1.21)) {//restrict scaling coefficients
1156	>	//if w is in the right range, so should be x, y, z.
1157	>	vector<StuntDouble*>::iterator sdi;
1158	>	Vector3d vel;
1159	>	for (sdi = coldBin.begin(); sdi != coldBin.end(); ++sdi) {
1160	>	if (rnemdFluxType_ == rnemdFullKE) {
1161	>	vel = (*sdi)->getVel() * c;
1162	>	(*sdi)->setVel(vel);
1163	>	}
1164	>	if ((*sdi)->isDirectional()) {
1165	>	Vector3d angMom = (*sdi)->getJ() * c;
1166	>	(*sdi)->setJ(angMom);
1167	>	}
1168	>	}
1169	>	w = sqrt(w);
1170	>	for (sdi = hotBin.begin(); sdi != hotBin.end(); ++sdi) {
1171	>	if (rnemdFluxType_ == rnemdFullKE) {
1172	>	vel = (*sdi)->getVel();
1173	>	vel.x() *= x;
1174	>	vel.y() *= y;
1175	>	vel.z() *= z;
1176	>	(*sdi)->setVel(vel);
1177	>	}
1178	>	if ((*sdi)->isDirectional()) {
1179	>	Vector3d angMom = (*sdi)->getJ() * w;
1180	>	(*sdi)->setJ(angMom);
1181	>	}
1182	>	}
1183	>	successfulScale = true;
1184	>	kineticExchange_ += kineticTarget_;
1185	>	}
1186	>	}
1187	>	} else {
1188	>	RealType a000, a110, c0, a001, a111, b01, b11, c1;
1189	>	switch(rnemdFluxType_) {
1190	>	case rnemdKE :
1191	>	/* used hotBin coeff's & only scale x & y dimensions
1192	>	RealType px = Phx / Pcx;
1193	>	RealType py = Phy / Pcy;
1194	>	a110 = Khy;
1195	>	c0 = - Khx - Khy - kineticTarget_;
1196	>	a000 = Khx;
1197	>	a111 = Kcy * py * py;
1198	>	b11 = -2.0 * Kcy * py * (1.0 + py);
1199	>	c1 = Kcy * py * (2.0 + py) + Kcx * px * ( 2.0 + px) + kineticTarget_;
1200	>	b01 = -2.0 * Kcx * px * (1.0 + px);
1201	>	a001 = Kcx * px * px;
1202	>	*/
1203	>	//scale all three dimensions, let c_x = c_y
1204	>	a000 = Kcx + Kcy;
1205	>	a110 = Kcz;
1206	>	c0 = kineticTarget_ - Kcx - Kcy - Kcz;
1207	>	a001 = Khx * px * px + Khy * py * py;
1208	>	a111 = Khz * pz * pz;
1209	>	b01 = -2.0 * (Khx * px * (1.0 + px) + Khy * py * (1.0 + py));
1210	>	b11 = -2.0 * Khz * pz * (1.0 + pz);
1211	>	c1 = Khx * px * (2.0 + px) + Khy * py * (2.0 + py)
1212	>	+ Khz * pz * (2.0 + pz) - kineticTarget_;
1213	>	break;
1214	>	case rnemdPx :
1215	>	c = 1 - momentumTarget_.x() / Pcx;
1216	>	a000 = Kcy;
1217	>	a110 = Kcz;
1218	>	c0 = Kcx * c * c - Kcx - Kcy - Kcz;
1219	>	a001 = py * py * Khy;
1220	>	a111 = pz * pz * Khz;
1221	>	b01 = -2.0 * Khy * py * (1.0 + py);
1222	>	b11 = -2.0 * Khz * pz * (1.0 + pz);
1223	>	c1 = Khy * py * (2.0 + py) + Khz * pz * (2.0 + pz)
1224	>	+ Khx * (fastpow(c * px - px - 1.0, 2) - 1.0);
1225	>	break;
1226	>	case rnemdPy :
1227	>	c = 1 - momentumTarget_.y() / Pcy;
1228	>	a000 = Kcx;
1229	>	a110 = Kcz;
1230	>	c0 = Kcy * c * c - Kcx - Kcy - Kcz;
1231	>	a001 = px * px * Khx;
1232	>	a111 = pz * pz * Khz;
1233	>	b01 = -2.0 * Khx * px * (1.0 + px);
1234	>	b11 = -2.0 * Khz * pz * (1.0 + pz);
1235	>	c1 = Khx * px * (2.0 + px) + Khz * pz * (2.0 + pz)
1236	>	+ Khy * (fastpow(c * py - py - 1.0, 2) - 1.0);
1237	>	break;
1238	>	case rnemdPz ://we don't really do this, do we?
1239	>	c = 1 - momentumTarget_.z() / Pcz;
1240	>	a000 = Kcx;
1241	>	a110 = Kcy;
1242	>	c0 = Kcz * c * c - Kcx - Kcy - Kcz;
1243	>	a001 = px * px * Khx;
1244	>	a111 = py * py * Khy;
1245	>	b01 = -2.0 * Khx * px * (1.0 + px);
1246	>	b11 = -2.0 * Khy * py * (1.0 + py);
1247	>	c1 = Khx * px * (2.0 + px) + Khy * py * (2.0 + py)
1248	>	+ Khz * (fastpow(c * pz - pz - 1.0, 2) - 1.0);
1249	>	break;
1250	>	default :
1251	>	break;
1252	>	}
1253	>
1254	>	RealType v1 = a000 * a111 - a001 * a110;
1255	>	RealType v2 = a000 * b01;
1256	>	RealType v3 = a000 * b11;
1257	>	RealType v4 = a000 * c1 - a001 * c0;
1258	>	RealType v8 = a110 * b01;
1259	>	RealType v10 = - b01 * c0;
1260	>
1261	>	RealType u0 = v2 * v10 - v4 * v4;
1262	>	RealType u1 = -2.0 * v3 * v4;
1263	>	RealType u2 = -v2 * v8 - v3 * v3 - 2.0 * v1 * v4;
1264	>	RealType u3 = -2.0 * v1 * v3;
1265	>	RealType u4 = - v1 * v1;
1266	>	//rescale coefficients
1267	>	RealType maxAbs = fabs(u0);
1268	>	if (maxAbs < fabs(u1)) maxAbs = fabs(u1);
1269	>	if (maxAbs < fabs(u2)) maxAbs = fabs(u2);
1270	>	if (maxAbs < fabs(u3)) maxAbs = fabs(u3);
1271	>	if (maxAbs < fabs(u4)) maxAbs = fabs(u4);
1272	>	u0 /= maxAbs;
1273	>	u1 /= maxAbs;
1274	>	u2 /= maxAbs;
1275	>	u3 /= maxAbs;
1276	>	u4 /= maxAbs;
1277	>	//max_element(start, end) is also available.
1278	>	Polynomial<RealType> poly; //same as DoublePolynomial poly;
1279	>	poly.setCoefficient(4, u4);
1280	>	poly.setCoefficient(3, u3);
1281	>	poly.setCoefficient(2, u2);
1282	>	poly.setCoefficient(1, u1);
1283	>	poly.setCoefficient(0, u0);
1284	>	vector<RealType> realRoots = poly.FindRealRoots();
1285	>
1286	>	vector<RealType>::iterator ri;
1287	>	RealType r1, r2, alpha0;
1288	>	vector<pair<RealType,RealType> > rps;
1289	>	for (ri = realRoots.begin(); ri !=realRoots.end(); ++ri) {
1290	>	r2 = *ri;
1291	>	//check if FindRealRoots() give the right answer
1292	>	if ( fabs(u0 + r2 * (u1 + r2 * (u2 + r2 * (u3 + r2 * u4)))) > 1e-6 ) {
1293	>	sprintf(painCave.errMsg,
1294	>	"RNEMD Warning: polynomial solve seems to have an error!");
1295	>	painCave.isFatal = 0;
1296	>	simError();
1297	>	failRootCount_++;
1298	>	}
1299	>	//might not be useful w/o rescaling coefficients
1300	>	alpha0 = -c0 - a110 * r2 * r2;
1301	>	if (alpha0 >= 0.0) {
1302	>	r1 = sqrt(alpha0 / a000);
1303	>	if (fabs(c1 + r1 * (b01 + r1 * a001) + r2 * (b11 + r2 * a111))
1304	>	< 1e-6)
1305	>	{ rps.push_back(make_pair(r1, r2)); }
1306	>	if (r1 > 1e-6) { //r1 non-negative
1307	>	r1 = -r1;
1308	>	if (fabs(c1 + r1 * (b01 + r1 * a001) + r2 * (b11 + r2 * a111))
1309	>	< 1e-6)
1310	>	{ rps.push_back(make_pair(r1, r2)); }
1311	>	}
1312	>	}
1313	>	}
1314	>	// Consider combining together the solving pair part w/ the searching
1315	>	// best solution part so that we don't need the pairs vector
1316	>	if (!rps.empty()) {
1317	>	RealType smallestDiff = HONKING_LARGE_VALUE;
1318	>	RealType diff;
1319	>	pair<RealType,RealType> bestPair = make_pair(1.0, 1.0);
1320	>	vector<pair<RealType,RealType> >::iterator rpi;
1321	>	for (rpi = rps.begin(); rpi != rps.end(); ++rpi) {
1322	>	r1 = (*rpi).first;
1323	>	r2 = (*rpi).second;
1324	>	switch(rnemdFluxType_) {
1325	>	case rnemdKE :
1326	>	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1327	>	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcx, 2)
1328	>	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcy, 2);
1329	>	break;
1330	>	case rnemdPx :
1331	>	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1332	>	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcy, 2);
1333	>	break;
1334	>	case rnemdPy :
1335	>	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1336	>	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcx, 2);
1337	>	break;
1338	>	case rnemdPz :
1339	>	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1340	>	+ fastpow(r1 * r1 / r2 / r2 - Kcy/Kcx, 2);
1341	>	default :
1342	>	break;
1343	>	}
1344	>	if (diff < smallestDiff) {
1345	>	smallestDiff = diff;
1346	>	bestPair = *rpi;
1347	>	}
1348	>	}
1349	>	#ifdef IS_MPI
1350	>	if (worldRank == 0) {
1351	>	#endif
1352	>	// sprintf(painCave.errMsg,
1353	>	//         "RNEMD: roots r1= %lf\tr2 = %lf\n",
1354	>	//         bestPair.first, bestPair.second);
1355	>	// painCave.isFatal = 0;
1356	>	// painCave.severity = OPENMD_INFO;
1357	>	// simError();
1358	>	#ifdef IS_MPI
1359	>	}
1360	>	#endif
1361	>
1362	>	switch(rnemdFluxType_) {
1363	>	case rnemdKE :
1364	>	x = bestPair.first;
1365	>	y = bestPair.first;
1366	>	z = bestPair.second;
1367	>	break;
1368	>	case rnemdPx :
1369	>	x = c;
1370	>	y = bestPair.first;
1371	>	z = bestPair.second;
1372	>	break;
1373	>	case rnemdPy :
1374	>	x = bestPair.first;
1375	>	y = c;
1376	>	z = bestPair.second;
1377	>	break;
1378	>	case rnemdPz :
1379	>	x = bestPair.first;
1380	>	y = bestPair.second;
1381	>	z = c;
1382	>	break;
1383	>	default :
1384	>	break;
1385	>	}
1386	>	vector<StuntDouble*>::iterator sdi;
1387	>	Vector3d vel;
1388	>	for (sdi = coldBin.begin(); sdi != coldBin.end(); ++sdi) {
1389	>	vel = (*sdi)->getVel();
1390	>	vel.x() *= x;
1391	>	vel.y() *= y;
1392	>	vel.z() *= z;
1393	>	(*sdi)->setVel(vel);
1394	>	}
1395	>	//convert to hotBin coefficient
1396	>	x = 1.0 + px * (1.0 - x);
1397	>	y = 1.0 + py * (1.0 - y);
1398	>	z = 1.0 + pz * (1.0 - z);
1399	>	for (sdi = hotBin.begin(); sdi != hotBin.end(); ++sdi) {
1400	>	vel = (*sdi)->getVel();
1401	>	vel.x() *= x;
1402	>	vel.y() *= y;
1403	>	vel.z() *= z;
1404	>	(*sdi)->setVel(vel);
1405	>	}
1406	>	successfulScale = true;
1407	>	switch(rnemdFluxType_) {
1408	>	case rnemdKE :
1409	>	kineticExchange_ += kineticTarget_;
1410	>	break;
1411	>	case rnemdPx :
1412	>	case rnemdPy :
1413	>	case rnemdPz :
1414	>	momentumExchange_ += momentumTarget_;
1415	>	break;
1416	>	default :
1417	>	break;
1418	>	}
1419	>	}
1420	>	}
1421	>	if (successfulScale != true) {
1422	>	sprintf(painCave.errMsg,
1423	>	"RNEMD::doNIVS exchange NOT performed - roots that solve\n"
1424	>	"\tthe constraint equations may not exist or there may be\n"
1425	>	"\tno selected objects in one or both slabs.\n");
1426	>	painCave.isFatal = 0;
1427	>	painCave.severity = OPENMD_INFO;
1428	>	simError();
1429	>	failTrialCount_++;
1430	>	}
1431	>	}
1432	>
1433	>	void RNEMD::doVSS(SelectionManager& smanA, SelectionManager& smanB) {
1434	>	if (!doRNEMD_) return;
1435	>	int selei;
1436	>	int selej;
1437	>
1438	>	Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
1439	>	RealType time = currentSnap_->getTime();
1440	>	Mat3x3d hmat = currentSnap_->getHmat();
1441	>
1442	>	StuntDouble* sd;
1443	>
1444	>	vector<StuntDouble*> hotBin, coldBin;
1445	>
1446	>	Vector3d Ph(V3Zero);
1447	>	Vector3d Lh(V3Zero);
1448	>	RealType Mh = 0.0;
1449	>	Mat3x3d Ih(0.0);
1450	>	RealType Kh = 0.0;
1451	>	Vector3d Pc(V3Zero);
1452	>	Vector3d Lc(V3Zero);
1453	>	RealType Mc = 0.0;
1454	>	Mat3x3d Ic(0.0);
1455	>	RealType Kc = 0.0;
1456	>
1457	>	// Constraints can be on only the linear or angular momentum, but
1458	>	// not both.  Usually, the user will specify which they want, but
1459	>	// in case they don't, the use of periodic boundaries should make
1460	>	// the choice for us.
1461	>	bool doLinearPart = false;
1462	>	bool doAngularPart = false;
1463	>
1464	>	switch (rnemdFluxType_) {
1465	>	case rnemdPx:
1466	>	case rnemdPy:
1467	>	case rnemdPz:
1468	>	case rnemdPvector:
1469	>	case rnemdKePx:
1470	>	case rnemdKePy:
1471	>	case rnemdKePvector:
1472	>	doLinearPart = true;
1473		break;
1474	<	case rnemdPzScale ://we don't really do this, do we?
1475	<	c = 1 - targetFlux_ / Pcz;
1476	<	a000 = Kcx;
1477	<	a110 = Kcy;
1478	<	c0 = Kcz * c * c - Kcx - Kcy - Kcz;
1479	<	a001 = px * px * Khx;
1480	<	a111 = py * py * Khy;
1481	<	b01 = -2.0 * Khx * px * (1.0 + px);
1482	<	b11 = -2.0 * Khy * py * (1.0 + py);
689	<	c1 = Khx * px * (2.0 + px) + Khy * py * (2.0 + py)
690	<	+ Khz * (fastpow(c * pz - pz - 1.0, 2) - 1.0);
691	<	break;
692	<	default :
1474	>	case rnemdLx:
1475	>	case rnemdLy:
1476	>	case rnemdLz:
1477	>	case rnemdLvector:
1478	>	case rnemdKeLx:
1479	>	case rnemdKeLy:
1480	>	case rnemdKeLz:
1481	>	case rnemdKeLvector:
1482	>	doAngularPart = true;
1483		break;
1484	+	case rnemdKE:
1485	+	case rnemdRotKE:
1486	+	case rnemdFullKE:
1487	+	default:
1488	+	if (usePeriodicBoundaryConditions_)
1489	+	doLinearPart = true;
1490	+	else
1491	+	doAngularPart = true;
1492	+	break;
1493		}
1494	+
1495	+	for (sd = smanA.beginSelected(selei); sd != NULL;
1496	+	sd = smanA.nextSelected(selei)) {
1497
1498	<	RealType v1 = a000 * a111 - a001 * a110;
697	<	RealType v2 = a000 * b01;
698	<	RealType v3 = a000 * b11;
699	<	RealType v4 = a000 * c1 - a001 * c0;
700	<	RealType v8 = a110 * b01;
701	<	RealType v10 = - b01 * c0;
1498	>	Vector3d pos = sd->getPos();
1499
1500	<	RealType u0 = v2 * v10 - v4 * v4;
1501	<	RealType u1 = -2.0 * v3 * v4;
1502	<	RealType u2 = -v2 * v8 - v3 * v3 - 2.0 * v1 * v4;
1503	<	RealType u3 = -2.0 * v1 * v3;
1504	<	RealType u4 = - v1 * v1;
1505	<	//rescale coefficients
1506	<	RealType maxAbs = fabs(u0);
1507	<	if (maxAbs < fabs(u1)) maxAbs = fabs(u1);
1508	<	if (maxAbs < fabs(u2)) maxAbs = fabs(u2);
1509	<	if (maxAbs < fabs(u3)) maxAbs = fabs(u3);
1510	<	if (maxAbs < fabs(u4)) maxAbs = fabs(u4);
1511	<	u0 /= maxAbs;
1512	<	u1 /= maxAbs;
1513	<	u2 /= maxAbs;
1514	<	u3 /= maxAbs;
1515	<	u4 /= maxAbs;
1516	<	//max_element(start, end) is also available.
1517	<	Polynomial<RealType> poly; //same as DoublePolynomial poly;
1518	<	poly.setCoefficient(4, u4);
1519	<	poly.setCoefficient(3, u3);
1520	<	poly.setCoefficient(2, u2);
1521	<	poly.setCoefficient(1, u1);
1522	<	poly.setCoefficient(0, u0);
1523	<	std::vector<RealType> realRoots = poly.FindRealRoots();
1524	<
1525	<	std::vector<RealType>::iterator ri;
1526	<	RealType r1, r2, alpha0;
1527	<	std::vector<std::pair<RealType,RealType> > rps;
1528	<	for (ri = realRoots.begin(); ri !=realRoots.end(); ri++) {
1529	<	r2 = *ri;
1530	<	//check if FindRealRoots() give the right answer
1531	<	if ( fabs(u0 + r2 * (u1 + r2 * (u2 + r2 * (u3 + r2 * u4)))) > 1e-6 ) {
1532	<	sprintf(painCave.errMsg,
1533	<	"RNEMD Warning: polynomial solve seems to have an error!");
1534	<	painCave.isFatal = 0;
1535	<	simError();
739	<	failRootCount_++;
740	<	}
741	<	//might not be useful w/o rescaling coefficients
742	<	alpha0 = -c0 - a110 * r2 * r2;
743	<	if (alpha0 >= 0.0) {
744	<	r1 = sqrt(alpha0 / a000);
745	<	if (fabs(c1 + r1 * (b01 + r1 * a001) + r2 * (b11 + r2 * a111)) < 1e-6)
746	<	{ rps.push_back(std::make_pair(r1, r2)); }
747	<	if (r1 > 1e-6) { //r1 non-negative
748	<	r1 = -r1;
749	<	if (fabs(c1 + r1 * (b01 + r1 * a001) + r2 * (b11 + r2 * a111)) <1e-6)
750	<	{ rps.push_back(std::make_pair(r1, r2)); }
1500	>	// wrap the stuntdouble's position back into the box:
1501	>
1502	>	if (usePeriodicBoundaryConditions_)
1503	>	currentSnap_->wrapVector(pos);
1504	>
1505	>	RealType mass = sd->getMass();
1506	>	Vector3d vel = sd->getVel();
1507	>	Vector3d rPos = sd->getPos() - coordinateOrigin_;
1508	>	RealType r2;
1509	>
1510	>	hotBin.push_back(sd);
1511	>	Ph += mass * vel;
1512	>	Mh += mass;
1513	>	Kh += mass * vel.lengthSquare();
1514	>	Lh += mass * cross(rPos, vel);
1515	>	Ih -= outProduct(rPos, rPos) * mass;
1516	>	r2 = rPos.lengthSquare();
1517	>	Ih(0, 0) += mass * r2;
1518	>	Ih(1, 1) += mass * r2;
1519	>	Ih(2, 2) += mass * r2;
1520	>
1521	>	if (rnemdFluxType_ == rnemdFullKE) {
1522	>	if (sd->isDirectional()) {
1523	>	Vector3d angMom = sd->getJ();
1524	>	Mat3x3d I = sd->getI();
1525	>	if (sd->isLinear()) {
1526	>	int i = sd->linearAxis();
1527	>	int j = (i + 1) % 3;
1528	>	int k = (i + 2) % 3;
1529	>	Kh += angMom[j] * angMom[j] / I(j, j) +
1530	>	angMom[k] * angMom[k] / I(k, k);
1531	>	} else {
1532	>	Kh += angMom[0] * angMom[0] / I(0, 0) +
1533	>	angMom[1] * angMom[1] / I(1, 1) +
1534	>	angMom[2] * angMom[2] / I(2, 2);
1535	>	}
1536		}
1537		}
1538		}
1539	<	// Consider combininig together the solving pair part w/ the searching
1540	<	// best solution part so that we don't need the pairs vector
1541	<	if (!rps.empty()) {
1542	<	RealType smallestDiff = HONKING_LARGE_VALUE;
1543	<	RealType diff;
1544	<	std::pair<RealType,RealType> bestPair = std::make_pair(1.0, 1.0);
1545	<	std::vector<std::pair<RealType,RealType> >::iterator rpi;
1546	<	for (rpi = rps.begin(); rpi != rps.end(); rpi++) {
1547	<	r1 = (*rpi).first;
1548	<	r2 = (*rpi).second;
1549	<	switch(rnemdType_) {
1550	<	case rnemdKineticScale :
1551	<	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1552	<	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcx, 2)
1553	<	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcy, 2);
1554	<	break;
1555	<	case rnemdPxScale :
1556	<	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1557	<	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcy, 2);
1558	<	break;
1559	<	case rnemdPyScale :
1560	<	diff = fastpow(1.0 - r1, 2) + fastpow(1.0 - r2, 2)
1561	<	+ fastpow(r1 * r1 / r2 / r2 - Kcz/Kcx, 2);
1562	<	break;
1563	<	case rnemdPzScale :
1564	<	default :
1565	<	break;
1539	>	for (sd = smanB.beginSelected(selej); sd != NULL;
1540	>	sd = smanB.nextSelected(selej)) {
1541	>
1542	>	Vector3d pos = sd->getPos();
1543	>
1544	>	// wrap the stuntdouble's position back into the box:
1545	>
1546	>	if (usePeriodicBoundaryConditions_)
1547	>	currentSnap_->wrapVector(pos);
1548	>
1549	>	RealType mass = sd->getMass();
1550	>	Vector3d vel = sd->getVel();
1551	>	Vector3d rPos = sd->getPos() - coordinateOrigin_;
1552	>	RealType r2;
1553	>
1554	>	coldBin.push_back(sd);
1555	>	Pc += mass * vel;
1556	>	Mc += mass;
1557	>	Kc += mass * vel.lengthSquare();
1558	>	Lc += mass * cross(rPos, vel);
1559	>	Ic -= outProduct(rPos, rPos) * mass;
1560	>	r2 = rPos.lengthSquare();
1561	>	Ic(0, 0) += mass * r2;
1562	>	Ic(1, 1) += mass * r2;
1563	>	Ic(2, 2) += mass * r2;
1564	>
1565	>	if (rnemdFluxType_ == rnemdFullKE) {
1566	>	if (sd->isDirectional()) {
1567	>	Vector3d angMom = sd->getJ();
1568	>	Mat3x3d I = sd->getI();
1569	>	if (sd->isLinear()) {
1570	>	int i = sd->linearAxis();
1571	>	int j = (i + 1) % 3;
1572	>	int k = (i + 2) % 3;
1573	>	Kc += angMom[j] * angMom[j] / I(j, j) +
1574	>	angMom[k] * angMom[k] / I(k, k);
1575	>	} else {
1576	>	Kc += angMom[0] * angMom[0] / I(0, 0) +
1577	>	angMom[1] * angMom[1] / I(1, 1) +
1578	>	angMom[2] * angMom[2] / I(2, 2);
1579	>	}
1580		}
782	–	if (diff < smallestDiff) {
783	–	smallestDiff = diff;
784	–	bestPair = *rpi;
785	–	}
1581		}
1582	+	}
1583	+
1584	+	Kh *= 0.5;
1585	+	Kc *= 0.5;
1586	+
1587		#ifdef IS_MPI
1588	<	if (worldRank == 0) {
1588	>	MPI_Allreduce(MPI_IN_PLACE, &Ph[0], 3, MPI_REALTYPE, MPI_SUM,
1589	>	MPI_COMM_WORLD);
1590	>	MPI_Allreduce(MPI_IN_PLACE, &Pc[0], 3, MPI_REALTYPE, MPI_SUM,
1591	>	MPI_COMM_WORLD);
1592	>	MPI_Allreduce(MPI_IN_PLACE, &Lh[0], 3, MPI_REALTYPE, MPI_SUM,
1593	>	MPI_COMM_WORLD);
1594	>	MPI_Allreduce(MPI_IN_PLACE, &Lc[0], 3, MPI_REALTYPE, MPI_SUM,
1595	>	MPI_COMM_WORLD);
1596	>	MPI_Allreduce(MPI_IN_PLACE, &Mh, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1597	>	MPI_Allreduce(MPI_IN_PLACE, &Kh, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1598	>	MPI_Allreduce(MPI_IN_PLACE, &Mc, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1599	>	MPI_Allreduce(MPI_IN_PLACE, &Kc, 1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1600	>	MPI_Allreduce(MPI_IN_PLACE, Ih.getArrayPointer(), 9,
1601	>	MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1602	>	MPI_Allreduce(MPI_IN_PLACE, Ic.getArrayPointer(), 9,
1603	>	MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
1604		#endif
1605	<	std::cerr << "we choose r1 = " << bestPair.first
791	<	<< " and r2 = " << bestPair.second << "\n";
792	<	#ifdef IS_MPI
793	<	}
794	<	#endif
1605	>
1606
1607	<	RealType x, y, z;
1608	<	switch(rnemdType_) {
1609	<	case rnemdKineticScale :
1610	<	x = bestPair.first;
1611	<	y = bestPair.first;
1612	<	z = bestPair.second;
1613	<	break;
1614	<	case rnemdPxScale :
1615	<	x = c;
1616	<	y = bestPair.first;
1617	<	z = bestPair.second;
1618	<	break;
1619	<	case rnemdPyScale :
1620	<	x = bestPair.first;
1621	<	y = c;
1622	<	z = bestPair.second;
1623	<	break;
1624	<	case rnemdPzScale :
1625	<	x = bestPair.first;
1626	<	y = bestPair.second;
1627	<	z = c;
1628	<	break;
1629	<	default :
1630	<	break;
1631	<	}
1632	<	std::vector<StuntDouble*>::iterator sdi;
1633	<	Vector3d vel;
1634	<	for (sdi = coldBin.begin(); sdi != coldBin.end(); sdi++) {
1635	<	vel = (*sdi)->getVel();
1636	<	vel.x() *= x;
1637	<	vel.y() *= y;
1638	<	vel.z() *= z;
1639	<	(*sdi)->setVel(vel);
1640	<	}
1641	<	//convert to hotBin coefficient
1642	<	x = 1.0 + px * (1.0 - x);
1643	<	y = 1.0 + py * (1.0 - y);
1644	<	z = 1.0 + pz * (1.0 - z);
1645	<	for (sdi = hotBin.begin(); sdi != hotBin.end(); sdi++) {
1646	<	vel = (*sdi)->getVel();
1647	<	vel.x() *= x;
1648	<	vel.y() *= y;
1649	<	vel.z() *= z;
1650	<	(*sdi)->setVel(vel);
1607	>	Vector3d ac, acrec, bc, bcrec;
1608	>	Vector3d ah, ahrec, bh, bhrec;
1609	>
1610	>	bool successfulExchange = false;
1611	>	if ((Mh > 0.0) && (Mc > 0.0)) {//both slabs are not empty
1612	>	Vector3d vc = Pc / Mc;
1613	>	ac = -momentumTarget_ / Mc + vc;
1614	>	acrec = -momentumTarget_ / Mc;
1615	>
1616	>	// We now need the inverse of the inertia tensor to calculate the
1617	>	// angular velocity of the cold slab;
1618	>	Mat3x3d Ici = Ic.inverse();
1619	>	Vector3d omegac = Ici * Lc;
1620	>	bc  = -(Ici * angularMomentumTarget_) + omegac;
1621	>	bcrec = bc - omegac;
1622	>
1623	>	RealType cNumerator = Kc - kineticTarget_;
1624	>	if (doLinearPart)
1625	>	cNumerator -= 0.5 * Mc * ac.lengthSquare();
1626	>
1627	>	if (doAngularPart)
1628	>	cNumerator -= 0.5 * ( dot(bc, Ic * bc));
1629	>
1630	>	if (cNumerator > 0.0) {
1631	>
1632	>	RealType cDenominator = Kc;
1633	>
1634	>	if (doLinearPart)
1635	>	cDenominator -= 0.5 * Mc * vc.lengthSquare();
1636	>
1637	>	if (doAngularPart)
1638	>	cDenominator -= 0.5*(dot(omegac, Ic * omegac));
1639	>
1640	>	if (cDenominator > 0.0) {
1641	>	RealType c = sqrt(cNumerator / cDenominator);
1642	>	if ((c > 0.9) && (c < 1.1)) {//restrict scaling coefficients
1643	>
1644	>	Vector3d vh = Ph / Mh;
1645	>	ah = momentumTarget_ / Mh + vh;
1646	>	ahrec = momentumTarget_ / Mh;
1647	>
1648	>	// We now need the inverse of the inertia tensor to
1649	>	// calculate the angular velocity of the hot slab;
1650	>	Mat3x3d Ihi = Ih.inverse();
1651	>	Vector3d omegah = Ihi * Lh;
1652	>	bh  = (Ihi * angularMomentumTarget_) + omegah;
1653	>	bhrec = bh - omegah;
1654	>
1655	>	RealType hNumerator = Kh + kineticTarget_;
1656	>	if (doLinearPart)
1657	>	hNumerator -= 0.5 * Mh * ah.lengthSquare();
1658	>
1659	>	if (doAngularPart)
1660	>	hNumerator -= 0.5 * ( dot(bh, Ih * bh));
1661	>
1662	>	if (hNumerator > 0.0) {
1663	>
1664	>	RealType hDenominator = Kh;
1665	>	if (doLinearPart)
1666	>	hDenominator -= 0.5 * Mh * vh.lengthSquare();
1667	>	if (doAngularPart)
1668	>	hDenominator -= 0.5*(dot(omegah, Ih * omegah));
1669	>
1670	>	if (hDenominator > 0.0) {
1671	>	RealType h = sqrt(hNumerator / hDenominator);
1672	>	if ((h > 0.9) && (h < 1.1)) {
1673	>
1674	>	vector<StuntDouble*>::iterator sdi;
1675	>	Vector3d vel;
1676	>	Vector3d rPos;
1677	>
1678	>	for (sdi = coldBin.begin(); sdi != coldBin.end(); ++sdi) {
1679	>	//vel = (*sdi)->getVel();
1680	>	rPos = (*sdi)->getPos() - coordinateOrigin_;
1681	>	if (doLinearPart)
1682	>	vel = ((*sdi)->getVel() - vc) * c + ac;
1683	>	if (doAngularPart)
1684	>	vel = ((*sdi)->getVel() - cross(omegac, rPos)) * c + cross(bc, rPos);
1685	>
1686	>	(*sdi)->setVel(vel);
1687	>	if (rnemdFluxType_ == rnemdFullKE) {
1688	>	if ((*sdi)->isDirectional()) {
1689	>	Vector3d angMom = (*sdi)->getJ() * c;
1690	>	(*sdi)->setJ(angMom);
1691	>	}
1692	>	}
1693	>	}
1694	>	for (sdi = hotBin.begin(); sdi != hotBin.end(); ++sdi) {
1695	>	//vel = (*sdi)->getVel();
1696	>	rPos = (*sdi)->getPos() - coordinateOrigin_;
1697	>	if (doLinearPart)
1698	>	vel = ((*sdi)->getVel() - vh) * h + ah;
1699	>	if (doAngularPart)
1700	>	vel = ((*sdi)->getVel() - cross(omegah, rPos)) * h + cross(bh, rPos);
1701	>
1702	>	(*sdi)->setVel(vel);
1703	>	if (rnemdFluxType_ == rnemdFullKE) {
1704	>	if ((*sdi)->isDirectional()) {
1705	>	Vector3d angMom = (*sdi)->getJ() * h;
1706	>	(*sdi)->setJ(angMom);
1707	>	}
1708	>	}
1709	>	}
1710	>	successfulExchange = true;
1711	>	kineticExchange_ += kineticTarget_;
1712	>	momentumExchange_ += momentumTarget_;
1713	>	angularMomentumExchange_ += angularMomentumTarget_;
1714	>	}
1715	>	}
1716	>	}
1717	>	}
1718	>	}
1719		}
1720	<	exchangeSum_ += targetFlux_;
1721	<	//we may want to check whether the exchange has been successful
1722	<	} else {
1723	<	std::cerr << "exchange NOT performed!\n";//MPI incompatible
1720	>	}
1721	>	if (successfulExchange != true) {
1722	>	sprintf(painCave.errMsg,
1723	>	"RNEMD::doVSS exchange NOT performed - roots that solve\n"
1724	>	"\tthe constraint equations may not exist or there may be\n"
1725	>	"\tno selected objects in one or both slabs.\n");
1726	>	painCave.isFatal = 0;
1727	>	painCave.severity = OPENMD_INFO;
1728	>	simError();
1729		failTrialCount_++;
1730		}
847	–
1731		}
1732
1733	+	RealType RNEMD::getDividingArea() {
1734	+
1735	+	if (hasDividingArea_) return dividingArea_;
1736	+
1737	+	RealType areaA, areaB;
1738	+	Snapshot* snap = info_->getSnapshotManager()->getCurrentSnapshot();
1739	+
1740	+	if (hasSelectionA_) {
1741	+
1742	+	if (evaluatorA_.hasSurfaceArea())
1743	+	areaA = evaluatorA_.getSurfaceArea();
1744	+	else {
1745	+
1746	+	cerr << "selection A did not have surface area, recomputing\n";
1747	+	int isd;
1748	+	StuntDouble* sd;
1749	+	vector<StuntDouble*> aSites;
1750	+	seleManA_.setSelectionSet(evaluatorA_.evaluate());
1751	+	for (sd = seleManA_.beginSelected(isd); sd != NULL;
1752	+	sd = seleManA_.nextSelected(isd)) {
1753	+	aSites.push_back(sd);
1754	+	}
1755	+	#if defined(HAVE_QHULL)
1756	+	ConvexHull* surfaceMeshA = new ConvexHull();
1757	+	surfaceMeshA->computeHull(aSites);
1758	+	areaA = surfaceMeshA->getArea();
1759	+	delete surfaceMeshA;
1760	+	#else
1761	+	sprintf( painCave.errMsg,
1762	+	"RNEMD::getDividingArea : Hull calculation is not possible\n"
1763	+	"\twithout libqhull. Please rebuild OpenMD with qhull enabled.");
1764	+	painCave.severity = OPENMD_ERROR;
1765	+	painCave.isFatal = 1;
1766	+	simError();
1767	+	#endif
1768	+	}
1769	+
1770	+	} else {
1771	+	if (usePeriodicBoundaryConditions_) {
1772	+	// in periodic boundaries, the surface area is twice the x-y
1773	+	// area of the current box:
1774	+	areaA = 2.0 * snap->getXYarea();
1775	+	} else {
1776	+	// in non-periodic simulations, without explicitly setting
1777	+	// selections, the sphere radius sets the surface area of the
1778	+	// dividing surface:
1779	+	areaA = 4.0 * M_PI * pow(sphereARadius_, 2);
1780	+	}
1781	+	}
1782	+
1783	+	if (hasSelectionB_) {
1784	+	if (evaluatorB_.hasSurfaceArea())
1785	+	areaB = evaluatorB_.getSurfaceArea();
1786	+	else {
1787	+	cerr << "selection B did not have surface area, recomputing\n";
1788	+
1789	+	int isd;
1790	+	StuntDouble* sd;
1791	+	vector<StuntDouble*> bSites;
1792	+	seleManB_.setSelectionSet(evaluatorB_.evaluate());
1793	+	for (sd = seleManB_.beginSelected(isd); sd != NULL;
1794	+	sd = seleManB_.nextSelected(isd)) {
1795	+	bSites.push_back(sd);
1796	+	}
1797	+
1798	+	#if defined(HAVE_QHULL)
1799	+	ConvexHull* surfaceMeshB = new ConvexHull();
1800	+	surfaceMeshB->computeHull(bSites);
1801	+	areaB = surfaceMeshB->getArea();
1802	+	delete surfaceMeshB;
1803	+	#else
1804	+	sprintf( painCave.errMsg,
1805	+	"RNEMD::getDividingArea : Hull calculation is not possible\n"
1806	+	"\twithout libqhull. Please rebuild OpenMD with qhull enabled.");
1807	+	painCave.severity = OPENMD_ERROR;
1808	+	painCave.isFatal = 1;
1809	+	simError();
1810	+	#endif
1811	+	}
1812	+
1813	+	} else {
1814	+	if (usePeriodicBoundaryConditions_) {
1815	+	// in periodic boundaries, the surface area is twice the x-y
1816	+	// area of the current box:
1817	+	areaB = 2.0 * snap->getXYarea();
1818	+	} else {
1819	+	// in non-periodic simulations, without explicitly setting
1820	+	// selections, but if a sphereBradius has been set, just use that:
1821	+	areaB = 4.0 * M_PI * pow(sphereBRadius_, 2);
1822	+	}
1823	+	}
1824	+
1825	+	dividingArea_ = min(areaA, areaB);
1826	+	hasDividingArea_ = true;
1827	+	return dividingArea_;
1828	+	}
1829	+
1830		void RNEMD::doRNEMD() {
1831	+	if (!doRNEMD_) return;
1832	+	trialCount_++;
1833
1834	<	switch(rnemdType_) {
1835	<	case rnemdKineticScale :
1836	<	case rnemdPxScale :
1837	<	case rnemdPyScale :
1838	<	case rnemdPzScale :
1839	<	doScale();
1834	>	// object evaluator:
1835	>	evaluator_.loadScriptString(rnemdObjectSelection_);
1836	>	seleMan_.setSelectionSet(evaluator_.evaluate());
1837	>
1838	>	evaluatorA_.loadScriptString(selectionA_);
1839	>	evaluatorB_.loadScriptString(selectionB_);
1840	>
1841	>	seleManA_.setSelectionSet(evaluatorA_.evaluate());
1842	>	seleManB_.setSelectionSet(evaluatorB_.evaluate());
1843	>
1844	>	commonA_ = seleManA_ & seleMan_;
1845	>	commonB_ = seleManB_ & seleMan_;
1846	>
1847	>	// Target exchange quantities (in each exchange) = dividingArea * dt * flux
1848	>	// dt = exchange time interval
1849	>	// flux = target flux
1850	>	// dividingArea = smallest dividing surface between the two regions
1851	>
1852	>	hasDividingArea_ = false;
1853	>	RealType area = getDividingArea();
1854	>
1855	>	kineticTarget_ = kineticFlux_ * exchangeTime_ * area;
1856	>	momentumTarget_ = momentumFluxVector_ * exchangeTime_ * area;
1857	>	angularMomentumTarget_ = angularMomentumFluxVector_ * exchangeTime_ * area;
1858	>
1859	>	switch(rnemdMethod_) {
1860	>	case rnemdSwap:
1861	>	doSwap(commonA_, commonB_);
1862		break;
1863	<	case rnemdKineticSwap :
1864	<	case rnemdPx :
861	<	case rnemdPy :
862	<	case rnemdPz :
863	<	doSwap();
1863	>	case rnemdNIVS:
1864	>	doNIVS(commonA_, commonB_);
1865		break;
1866	<	case rnemdUnknown :
1866	>	case rnemdVSS:
1867	>	doVSS(commonA_, commonB_);
1868	>	break;
1869	>	case rnemdUnkownMethod:
1870		default :
1871		break;
1872		}
1873		}
1874
1875		void RNEMD::collectData() {
1876	<
1876	>	if (!doRNEMD_) return;
1877		Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
1878	+
1879	+	// collectData can be called more frequently than the doRNEMD, so use the
1880	+	// computed area from the last exchange time:
1881	+	RealType area = getDividingArea();
1882	+	areaAccumulator_->add(area);
1883		Mat3x3d hmat = currentSnap_->getHmat();
1884	+	Vector3d u = angularMomentumFluxVector_;
1885	+	u.normalize();
1886
1887		seleMan_.setSelectionSet(evaluator_.evaluate());
1888
1889	<	int selei;
1889	>	int selei(0);
1890		StuntDouble* sd;
1891	<	int idx;
1891	>	int binNo;
1892	>	RealType mass;
1893	>	Vector3d vel;
1894	>	Vector3d rPos;
1895	>	RealType KE;
1896	>	Vector3d L;
1897	>	Mat3x3d I;
1898	>	RealType r2;
1899
1900	+	vector<RealType> binMass(nBins_, 0.0);
1901	+	vector<Vector3d> binP(nBins_, V3Zero);
1902	+	vector<RealType> binOmega(nBins_, 0.0);
1903	+	vector<Vector3d> binL(nBins_, V3Zero);
1904	+	vector<Mat3x3d>  binI(nBins_);
1905	+	vector<RealType> binKE(nBins_, 0.0);
1906	+	vector<int> binDOF(nBins_, 0);
1907	+	vector<int> binCount(nBins_, 0);
1908	+
1909	+	// alternative approach, track all molecules instead of only those
1910	+	// selected for scaling/swapping:
1911	+	/*
1912	+	SimInfo::MoleculeIterator miter;
1913	+	vector<StuntDouble*>::iterator iiter;
1914	+	Molecule* mol;
1915	+	StuntDouble* sd;
1916	+	for (mol = info_->beginMolecule(miter); mol != NULL;
1917	+	mol = info_->nextMolecule(miter))
1918	+	sd is essentially sd
1919	+	for (sd = mol->beginIntegrableObject(iiter);
1920	+	sd != NULL;
1921	+	sd = mol->nextIntegrableObject(iiter))
1922	+	*/
1923	+
1924		for (sd = seleMan_.beginSelected(selei); sd != NULL;
1925	<	sd = seleMan_.nextSelected(selei)) {
1926	<
885	<	idx = sd->getLocalIndex();
886	<
1925	>	sd = seleMan_.nextSelected(selei)) {
1926	>
1927		Vector3d pos = sd->getPos();
1928
1929		// wrap the stuntdouble's position back into the box:
1930
1931	<	if (usePeriodicBoundaryConditions_)
1931	>	if (usePeriodicBoundaryConditions_) {
1932		currentSnap_->wrapVector(pos);
1933	<
1934	<	// which bin is this stuntdouble in?
1935	<	// wrapped positions are in the range [-0.5*hmat(2,2), +0.5*hmat(2,2)]
1936	<
1937	<	int binNo = int(nBins_ * (pos.z() / hmat(2,2) + 0.5)) % nBins_;
1933	>	// which bin is this stuntdouble in?
1934	>	// wrapped positions are in the range [-0.5*hmat(2,2), +0.5*hmat(2,2)]
1935	>	// Shift molecules by half a box to have bins start at 0
1936	>	// The modulo operator is used to wrap the case when we are
1937	>	// beyond the end of the bins back to the beginning.
1938	>	binNo = int(nBins_ * (pos.z() / hmat(2,2) + 0.5)) % nBins_;
1939	>	} else {
1940	>	Vector3d rPos = pos - coordinateOrigin_;
1941	>	binNo = int(rPos.length() / binWidth_);
1942	>	}
1943
1944	<	if (rnemdLogWidth_ == midBin_ + 1)
1945	<	if (binNo > midBin_)
1946	<	binNo = nBins_ - binNo;
1944	>	mass = sd->getMass();
1945	>	vel = sd->getVel();
1946	>	rPos = sd->getPos() - coordinateOrigin_;
1947	>	KE = 0.5 * mass * vel.lengthSquare();
1948	>	L = mass * cross(rPos, vel);
1949	>	I = outProduct(rPos, rPos) * mass;
1950	>	r2 = rPos.lengthSquare();
1951	>	I(0, 0) += mass * r2;
1952	>	I(1, 1) += mass * r2;
1953	>	I(2, 2) += mass * r2;
1954
1955	<	RealType mass = sd->getMass();
1956	<	Vector3d vel = sd->getVel();
1957	<	RealType value;
1958	<	RealType xVal, yVal, zVal;
1955	>	// Project the relative position onto a plane perpendicular to
1956	>	// the angularMomentumFluxVector:
1957	>	// Vector3d rProj = rPos - dot(rPos, u) * u;
1958	>	// Project the velocity onto a plane perpendicular to the
1959	>	// angularMomentumFluxVector:
1960	>	// Vector3d vProj = vel  - dot(vel, u) * u;
1961	>	// Compute angular velocity vector (should be nearly parallel to
1962	>	// angularMomentumFluxVector
1963	>	// Vector3d aVel = cross(rProj, vProj);
1964
1965	<	switch(rnemdType_) {
1966	<	case rnemdKineticSwap :
1967	<	case rnemdKineticScale :
1968	<
1969	<	value = mass * (vel[0]*vel[0] + vel[1]*vel[1] +
1970	<	vel[2]*vel[2]);
1971	<
1972	<	valueCount_[binNo] += 3;
1973	<	if (sd->isDirectional()) {
1974	<	Vector3d angMom = sd->getJ();
1975	<	Mat3x3d I = sd->getI();
1976	<
1977	<	if (sd->isLinear()) {
1978	<	int i = sd->linearAxis();
1979	<	int j = (i + 1) % 3;
1980	<	int k = (i + 2) % 3;
1981	<	value += angMom[j] * angMom[j] / I(j, j) +
1982	<	angMom[k] * angMom[k] / I(k, k);
1983	<
1984	<	valueCount_[binNo] +=2;
1985	<
1986	<	} else {
1987	<	value += angMom[0]*angMom[0]/I(0, 0)
1988	<	+ angMom[1]*angMom[1]/I(1, 1)
1989	<	+ angMom[2]*angMom[2]/I(2, 2);
1990	<	valueCount_[binNo] +=3;
934	<	}
935	<	}
936	<	value = value / PhysicalConstants::energyConvert / PhysicalConstants::kb;
937	<
938	<	break;
939	<	case rnemdPx :
940	<	case rnemdPxScale :
941	<	value = mass * vel[0];
942	<	valueCount_[binNo]++;
943	<	xVal = mass * vel.x() * vel.x() / PhysicalConstants::energyConvert
944	<	/ PhysicalConstants::kb;
945	<	yVal = mass * vel.y() * vel.y() / PhysicalConstants::energyConvert
946	<	/ PhysicalConstants::kb;
947	<	zVal = mass * vel.z() * vel.z() / PhysicalConstants::energyConvert
948	<	/ PhysicalConstants::kb;
949	<	xTempHist_[binNo] += xVal;
950	<	yTempHist_[binNo] += yVal;
951	<	zTempHist_[binNo] += zVal;
952	<	break;
953	<	case rnemdPy :
954	<	case rnemdPyScale :
955	<	value = mass * vel[1];
956	<	valueCount_[binNo]++;
957	<	break;
958	<	case rnemdPz :
959	<	case rnemdPzScale :
960	<	value = mass * vel[2];
961	<	valueCount_[binNo]++;
962	<	break;
963	<	case rnemdUnknown :
964	<	default :
965	<	break;
1965	>	if (binNo >= 0 && binNo < nBins_)  {
1966	>	binCount[binNo]++;
1967	>	binMass[binNo] += mass;
1968	>	binP[binNo] += mass*vel;
1969	>	binKE[binNo] += KE;
1970	>	binI[binNo] += I;
1971	>	binL[binNo] += L;
1972	>	binDOF[binNo] += 3;
1973	>
1974	>	if (sd->isDirectional()) {
1975	>	Vector3d angMom = sd->getJ();
1976	>	Mat3x3d Ia = sd->getI();
1977	>	if (sd->isLinear()) {
1978	>	int i = sd->linearAxis();
1979	>	int j = (i + 1) % 3;
1980	>	int k = (i + 2) % 3;
1981	>	binKE[binNo] += 0.5 * (angMom[j] * angMom[j] / Ia(j, j) +
1982	>	angMom[k] * angMom[k] / Ia(k, k));
1983	>	binDOF[binNo] += 2;
1984	>	} else {
1985	>	binKE[binNo] += 0.5 * (angMom[0] * angMom[0] / Ia(0, 0) +
1986	>	angMom[1] * angMom[1] / Ia(1, 1) +
1987	>	angMom[2] * angMom[2] / Ia(2, 2));
1988	>	binDOF[binNo] += 3;
1989	>	}
1990	>	}
1991		}
1992	<	valueHist_[binNo] += value;
1992	>	}
1993	>
1994	>	#ifdef IS_MPI
1995	>
1996	>	for (int i = 0; i < nBins_; i++) {
1997	>
1998	>	MPI_Allreduce(MPI_IN_PLACE, &binCount[i],
1999	>	1, MPI_INT, MPI_SUM, MPI_COMM_WORLD);
2000	>	MPI_Allreduce(MPI_IN_PLACE, &binMass[i],
2001	>	1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2002	>	MPI_Allreduce(MPI_IN_PLACE, binP[i].getArrayPointer(),
2003	>	3, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2004	>	MPI_Allreduce(MPI_IN_PLACE, binL[i].getArrayPointer(),
2005	>	3, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2006	>	MPI_Allreduce(MPI_IN_PLACE, binI[i].getArrayPointer(),
2007	>	9, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2008	>	MPI_Allreduce(MPI_IN_PLACE, &binKE[i],
2009	>	1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2010	>	MPI_Allreduce(MPI_IN_PLACE, &binDOF[i],
2011	>	1, MPI_INT, MPI_SUM, MPI_COMM_WORLD);
2012	>	//MPI_Allreduce(MPI_IN_PLACE, &binOmega[i],
2013	>	//                          1, MPI_REALTYPE, MPI_SUM, MPI_COMM_WORLD);
2014		}
2015	+
2016	+	#endif
2017
2018	+	Vector3d omega;
2019	+	RealType den;
2020	+	RealType temp;
2021	+	RealType z;
2022	+	RealType r;
2023	+	for (int i = 0; i < nBins_; i++) {
2024	+	if (usePeriodicBoundaryConditions_) {
2025	+	z = (((RealType)i + 0.5) / (RealType)nBins_) * hmat(2,2);
2026	+	den = binMass[i] * nBins_ * PhysicalConstants::densityConvert
2027	+	/ currentSnap_->getVolume() ;
2028	+	} else {
2029	+	r = (((RealType)i + 0.5) * binWidth_);
2030	+	RealType rinner = (RealType)i * binWidth_;
2031	+	RealType router = (RealType)(i+1) * binWidth_;
2032	+	den = binMass[i] * 3.0 * PhysicalConstants::densityConvert
2033	+	/ (4.0 * M_PI * (pow(router,3) - pow(rinner,3)));
2034	+	}
2035	+	vel = binP[i] / binMass[i];
2036	+
2037	+	omega = binI[i].inverse() * binL[i];
2038	+
2039	+	// omega = binOmega[i] / binCount[i];
2040	+
2041	+	if (binCount[i] > 0) {
2042	+	// only add values if there are things to add
2043	+	temp = 2.0 * binKE[i] / (binDOF[i] * PhysicalConstants::kb *
2044	+	PhysicalConstants::energyConvert);
2045	+
2046	+	for (unsigned int j = 0; j < outputMask_.size(); ++j) {
2047	+	if(outputMask_[j]) {
2048	+	switch(j) {
2049	+	case Z:
2050	+	dynamic_cast<Accumulator *>(data_[j].accumulator[i])->add(z);
2051	+	break;
2052	+	case R:
2053	+	dynamic_cast<Accumulator *>(data_[j].accumulator[i])->add(r);
2054	+	break;
2055	+	case TEMPERATURE:
2056	+	dynamic_cast<Accumulator *>(data_[j].accumulator[i])->add(temp);
2057	+	break;
2058	+	case VELOCITY:
2059	+	dynamic_cast<VectorAccumulator *>(data_[j].accumulator[i])->add(vel);
2060	+	break;
2061	+	case ANGULARVELOCITY:
2062	+	dynamic_cast<VectorAccumulator *>(data_[j].accumulator[i])->add(omega);
2063	+	break;
2064	+	case DENSITY:
2065	+	dynamic_cast<Accumulator *>(data_[j].accumulator[i])->add(den);
2066	+	break;
2067	+	}
2068	+	}
2069	+	}
2070	+	}
2071	+	}
2072	+	hasData_ = true;
2073		}
2074
2075		void RNEMD::getStarted() {
2076	<	Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
2077	<	Stats& stat = currentSnap_->statData;
2078	<	stat[Stats::RNEMD_EXCHANGE_TOTAL] = exchangeSum_;
2076	>	if (!doRNEMD_) return;
2077	>	hasDividingArea_ = false;
2078	>	collectData();
2079	>	writeOutputFile();
2080		}
2081
2082	<	void RNEMD::getStatus() {
2082	>	void RNEMD::parseOutputFileFormat(const std::string& format) {
2083	>	if (!doRNEMD_) return;
2084	>	StringTokenizer tokenizer(format, " ,;|\t\n\r");
2085	>
2086	>	while(tokenizer.hasMoreTokens()) {
2087	>	std::string token(tokenizer.nextToken());
2088	>	toUpper(token);
2089	>	OutputMapType::iterator i = outputMap_.find(token);
2090	>	if (i != outputMap_.end()) {
2091	>	outputMask_.set(i->second);
2092	>	} else {
2093	>	sprintf( painCave.errMsg,
2094	>	"RNEMD::parseOutputFileFormat: %s is not a recognized\n"
2095	>	"\toutputFileFormat keyword.\n", token.c_str() );
2096	>	painCave.isFatal = 0;
2097	>	painCave.severity = OPENMD_ERROR;
2098	>	simError();
2099	>	}
2100	>	}
2101	>	}
2102	>
2103	>	void RNEMD::writeOutputFile() {
2104	>	if (!doRNEMD_) return;
2105	>	if (!hasData_) return;
2106	>
2107	>	#ifdef IS_MPI
2108	>	// If we're the root node, should we print out the results
2109	>	int worldRank;
2110	>	MPI_Comm_rank( MPI_COMM_WORLD, &worldRank);
2111
2112	<	Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
2113	<	Stats& stat = currentSnap_->statData;
2114	<	RealType time = currentSnap_->getTime();
2112	>	if (worldRank == 0) {
2113	>	#endif
2114	>	rnemdFile_.open(rnemdFileName_.c_str(), std::ios::out | std::ios::trunc );
2115	>
2116	>	if( !rnemdFile_ ){
2117	>	sprintf( painCave.errMsg,
2118	>	"Could not open \"%s\" for RNEMD output.\n",
2119	>	rnemdFileName_.c_str());
2120	>	painCave.isFatal = 1;
2121	>	simError();
2122	>	}
2123
2124	<	stat[Stats::RNEMD_EXCHANGE_TOTAL] = exchangeSum_;
985	<	//or to be more meaningful, define another item as exchangeSum_ / time
986	<	int j;
2124	>	Snapshot* currentSnap_ = info_->getSnapshotManager()->getCurrentSnapshot();
2125
2126	<	#ifdef IS_MPI
2126	>	RealType time = currentSnap_->getTime();
2127	>	RealType avgArea;
2128	>	areaAccumulator_->getAverage(avgArea);
2129
2130	<	// all processors have the same number of bins, and STL vectors pack their
2131	<	// arrays, so in theory, this should be safe:
2130	>	RealType Jz(0.0);
2131	>	Vector3d JzP(V3Zero);
2132	>	Vector3d JzL(V3Zero);
2133	>	if (time >= info_->getSimParams()->getDt()) {
2134	>	Jz = kineticExchange_ / (time * avgArea)
2135	>	/ PhysicalConstants::energyConvert;
2136	>	JzP = momentumExchange_ / (time * avgArea);
2137	>	JzL = angularMomentumExchange_ / (time * avgArea);
2138	>	}
2139
2140	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &valueHist_[0],
2141	<	rnemdLogWidth_, MPI::REALTYPE, MPI::SUM);
2142	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &valueCount_[0],
2143	<	rnemdLogWidth_, MPI::INT, MPI::SUM);
2144	<	if (rnemdType_ == rnemdPx || rnemdType_ == rnemdPxScale) {
2145	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &xTempHist_[0],
2146	<	rnemdLogWidth_, MPI::REALTYPE, MPI::SUM);
1000	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &yTempHist_[0],
1001	<	rnemdLogWidth_, MPI::REALTYPE, MPI::SUM);
1002	<	MPI::COMM_WORLD.Allreduce(MPI::IN_PLACE, &zTempHist_[0],
1003	<	rnemdLogWidth_, MPI::REALTYPE, MPI::SUM);
1004	<	}
1005	<	// If we're the root node, should we print out the results
1006	<	int worldRank = MPI::COMM_WORLD.Get_rank();
1007	<	if (worldRank == 0) {
1008	<	#endif
1009	<	rnemdLog_ << time;
1010	<	for (j = 0; j < rnemdLogWidth_; j++) {
1011	<	rnemdLog_ << "\t" << valueHist_[j] / (RealType)valueCount_[j];
2140	>	rnemdFile_ << "#######################################################\n";
2141	>	rnemdFile_ << "# RNEMD {\n";
2142	>
2143	>	map<string, RNEMDMethod>::iterator mi;
2144	>	for(mi = stringToMethod_.begin(); mi != stringToMethod_.end(); ++mi) {
2145	>	if ( (*mi).second == rnemdMethod_)
2146	>	rnemdFile_ << "#    exchangeMethod  = \"" << (*mi).first << "\";\n";
2147		}
2148	<	rnemdLog_ << "\n";
2149	<	if (rnemdType_ == rnemdPx || rnemdType_ == rnemdPxScale ) {
2150	<	xTempLog_ << time;
2151	<	for (j = 0; j < rnemdLogWidth_; j++) {
2152	<	xTempLog_ << "\t" << xTempHist_[j] / (RealType)valueCount_[j];
2148	>	map<string, RNEMDFluxType>::iterator fi;
2149	>	for(fi = stringToFluxType_.begin(); fi != stringToFluxType_.end(); ++fi) {
2150	>	if ( (*fi).second == rnemdFluxType_)
2151	>	rnemdFile_ << "#    fluxType  = \"" << (*fi).first << "\";\n";
2152	>	}
2153	>
2154	>	rnemdFile_ << "#    exchangeTime = " << exchangeTime_ << ";\n";
2155	>
2156	>	rnemdFile_ << "#    objectSelection = \""
2157	>	<< rnemdObjectSelection_ << "\";\n";
2158	>	rnemdFile_ << "#    selectionA = \"" << selectionA_ << "\";\n";
2159	>	rnemdFile_ << "#    selectionB = \"" << selectionB_ << "\";\n";
2160	>	rnemdFile_ << "# }\n";
2161	>	rnemdFile_ << "#######################################################\n";
2162	>	rnemdFile_ << "# RNEMD report:\n";
2163	>	rnemdFile_ << "#      running time = " << time << " fs\n";
2164	>	rnemdFile_ << "# Target flux:\n";
2165	>	rnemdFile_ << "#           kinetic = "
2166	>	<< kineticFlux_ / PhysicalConstants::energyConvert
2167	>	<< " (kcal/mol/A^2/fs)\n";
2168	>	rnemdFile_ << "#          momentum = " << momentumFluxVector_
2169	>	<< " (amu/A/fs^2)\n";
2170	>	rnemdFile_ << "#  angular momentum = " << angularMomentumFluxVector_
2171	>	<< " (amu/A^2/fs^2)\n";
2172	>	rnemdFile_ << "# Target one-time exchanges:\n";
2173	>	rnemdFile_ << "#          kinetic = "
2174	>	<< kineticTarget_ / PhysicalConstants::energyConvert
2175	>	<< " (kcal/mol)\n";
2176	>	rnemdFile_ << "#          momentum = " << momentumTarget_
2177	>	<< " (amu*A/fs)\n";
2178	>	rnemdFile_ << "#  angular momentum = " << angularMomentumTarget_
2179	>	<< " (amu*A^2/fs)\n";
2180	>	rnemdFile_ << "# Actual exchange totals:\n";
2181	>	rnemdFile_ << "#          kinetic = "
2182	>	<< kineticExchange_ / PhysicalConstants::energyConvert
2183	>	<< " (kcal/mol)\n";
2184	>	rnemdFile_ << "#          momentum = " << momentumExchange_
2185	>	<< " (amu*A/fs)\n";
2186	>	rnemdFile_ << "#  angular momentum = " << angularMomentumExchange_
2187	>	<< " (amu*A^2/fs)\n";
2188	>	rnemdFile_ << "# Actual flux:\n";
2189	>	rnemdFile_ << "#          kinetic = " << Jz
2190	>	<< " (kcal/mol/A^2/fs)\n";
2191	>	rnemdFile_ << "#          momentum = " << JzP
2192	>	<< " (amu/A/fs^2)\n";
2193	>	rnemdFile_ << "#  angular momentum = " << JzL
2194	>	<< " (amu/A^2/fs^2)\n";
2195	>	rnemdFile_ << "# Exchange statistics:\n";
2196	>	rnemdFile_ << "#               attempted = " << trialCount_ << "\n";
2197	>	rnemdFile_ << "#                  failed = " << failTrialCount_ << "\n";
2198	>	if (rnemdMethod_ == rnemdNIVS) {
2199	>	rnemdFile_ << "#  NIVS root-check errors = "
2200	>	<< failRootCount_ << "\n";
2201	>	}
2202	>	rnemdFile_ << "#######################################################\n";
2203	>
2204	>
2205	>
2206	>	//write title
2207	>	rnemdFile_ << "#";
2208	>	for (unsigned int i = 0; i < outputMask_.size(); ++i) {
2209	>	if (outputMask_[i]) {
2210	>	rnemdFile_ << "\t" << data_[i].title <<
2211	>	"(" << data_[i].units << ")";
2212	>	// add some extra tabs for column alignment
2213	>	if (data_[i].dataType == "Vector3d") rnemdFile_ << "\t\t";
2214		}
2215	<	xTempLog_ << "\n";
2216	<	yTempLog_ << time;
2217	<	for (j = 0; j < rnemdLogWidth_; j++) {
2218	<	yTempLog_ << "\t" << yTempHist_[j] / (RealType)valueCount_[j];
2215	>	}
2216	>	rnemdFile_ << std::endl;
2217	>
2218	>	rnemdFile_.precision(8);
2219	>
2220	>	for (int j = 0; j < nBins_; j++) {
2221	>
2222	>	for (unsigned int i = 0; i < outputMask_.size(); ++i) {
2223	>	if (outputMask_[i]) {
2224	>	if (data_[i].dataType == "RealType")
2225	>	writeReal(i,j);
2226	>	else if (data_[i].dataType == "Vector3d")
2227	>	writeVector(i,j);
2228	>	else {
2229	>	sprintf( painCave.errMsg,
2230	>	"RNEMD found an unknown data type for: %s ",
2231	>	data_[i].title.c_str());
2232	>	painCave.isFatal = 1;
2233	>	simError();
2234	>	}
2235	>	}
2236		}
2237	<	yTempLog_ << "\n";
2238	<	zTempLog_ << time;
2239	<	for (j = 0; j < rnemdLogWidth_; j++) {
2240	<	zTempLog_ << "\t" << zTempHist_[j] / (RealType)valueCount_[j];
2237	>	rnemdFile_ << std::endl;
2238	>
2239	>	}
2240	>
2241	>	rnemdFile_ << "#######################################################\n";
2242	>	rnemdFile_ << "# Standard Deviations in those quantities follow:\n";
2243	>	rnemdFile_ << "#######################################################\n";
2244	>
2245	>
2246	>	for (int j = 0; j < nBins_; j++) {
2247	>	rnemdFile_ << "#";
2248	>	for (unsigned int i = 0; i < outputMask_.size(); ++i) {
2249	>	if (outputMask_[i]) {
2250	>	if (data_[i].dataType == "RealType")
2251	>	writeRealStdDev(i,j);
2252	>	else if (data_[i].dataType == "Vector3d")
2253	>	writeVectorStdDev(i,j);
2254	>	else {
2255	>	sprintf( painCave.errMsg,
2256	>	"RNEMD found an unknown data type for: %s ",
2257	>	data_[i].title.c_str());
2258	>	painCave.isFatal = 1;
2259	>	simError();
2260	>	}
2261	>	}
2262		}
2263	<	zTempLog_ << "\n";
2264	<	}
2263	>	rnemdFile_ << std::endl;
2264	>
2265	>	}
2266	>
2267	>	rnemdFile_.flush();
2268	>	rnemdFile_.close();
2269	>
2270		#ifdef IS_MPI
2271		}
2272		#endif
2273	<	for (j = 0; j < rnemdLogWidth_; j++) {
2274	<	valueCount_[j] = 0;
2275	<	valueHist_[j] = 0.0;
2273	>
2274	>	}
2275	>
2276	>	void RNEMD::writeReal(int index, unsigned int bin) {
2277	>	if (!doRNEMD_) return;
2278	>	assert(index >=0 && index < ENDINDEX);
2279	>	assert(int(bin) < nBins_);
2280	>	RealType s;
2281	>	int count;
2282	>
2283	>	count = data_[index].accumulator[bin]->count();
2284	>	if (count == 0) return;
2285	>
2286	>	dynamic_cast<Accumulator *>(data_[index].accumulator[bin])->getAverage(s);
2287	>
2288	>	if (! isinf(s) && ! isnan(s)) {
2289	>	rnemdFile_ << "\t" << s;
2290	>	} else{
2291	>	sprintf( painCave.errMsg,
2292	>	"RNEMD detected a numerical error writing: %s for bin %u",
2293	>	data_[index].title.c_str(), bin);
2294	>	painCave.isFatal = 1;
2295	>	simError();
2296	>	}
2297	>	}
2298	>
2299	>	void RNEMD::writeVector(int index, unsigned int bin) {
2300	>	if (!doRNEMD_) return;
2301	>	assert(index >=0 && index < ENDINDEX);
2302	>	assert(int(bin) < nBins_);
2303	>	Vector3d s;
2304	>	int count;
2305	>
2306	>	count = data_[index].accumulator[bin]->count();
2307	>
2308	>	if (count == 0) return;
2309	>
2310	>	dynamic_cast<VectorAccumulator*>(data_[index].accumulator[bin])->getAverage(s);
2311	>	if (isinf(s[0]) || isnan(s[0]) ||
2312	>	isinf(s[1]) || isnan(s[1]) ||
2313	>	isinf(s[2]) || isnan(s[2]) ) {
2314	>	sprintf( painCave.errMsg,
2315	>	"RNEMD detected a numerical error writing: %s for bin %u",
2316	>	data_[index].title.c_str(), bin);
2317	>	painCave.isFatal = 1;
2318	>	simError();
2319	>	} else {
2320	>	rnemdFile_ << "\t" << s[0] << "\t" << s[1] << "\t" << s[2];
2321		}
2322	<	if (rnemdType_ == rnemdPx || rnemdType_ == rnemdPxScale)
2323	<	for (j = 0; j < rnemdLogWidth_; j++) {
2324	<	xTempHist_[j] = 0.0;
2325	<	yTempHist_[j] = 0.0;
2326	<	zTempHist_[j] = 0.0;
2327	<	}
2322	>	}
2323	>
2324	>	void RNEMD::writeRealStdDev(int index, unsigned int bin) {
2325	>	if (!doRNEMD_) return;
2326	>	assert(index >=0 && index < ENDINDEX);
2327	>	assert(int(bin) < nBins_);
2328	>	RealType s;
2329	>	int count;
2330	>
2331	>	count = data_[index].accumulator[bin]->count();
2332	>	if (count == 0) return;
2333	>
2334	>	dynamic_cast<Accumulator *>(data_[index].accumulator[bin])->getStdDev(s);
2335	>
2336	>	if (! isinf(s) && ! isnan(s)) {
2337	>	rnemdFile_ << "\t" << s;
2338	>	} else{
2339	>	sprintf( painCave.errMsg,
2340	>	"RNEMD detected a numerical error writing: %s std. dev. for bin %u",
2341	>	data_[index].title.c_str(), bin);
2342	>	painCave.isFatal = 1;
2343	>	simError();
2344	>	}
2345		}
2346	+
2347	+	void RNEMD::writeVectorStdDev(int index, unsigned int bin) {
2348	+	if (!doRNEMD_) return;
2349	+	assert(index >=0 && index < ENDINDEX);
2350	+	assert(int(bin) < nBins_);
2351	+	Vector3d s;
2352	+	int count;
2353	+
2354	+	count = data_[index].accumulator[bin]->count();
2355	+	if (count == 0) return;
2356	+
2357	+	dynamic_cast<VectorAccumulator*>(data_[index].accumulator[bin])->getStdDev(s);
2358	+	if (isinf(s[0]) || isnan(s[0]) ||
2359	+	isinf(s[1]) || isnan(s[1]) ||
2360	+	isinf(s[2]) || isnan(s[2]) ) {
2361	+	sprintf( painCave.errMsg,
2362	+	"RNEMD detected a numerical error writing: %s std. dev. for bin %u",
2363	+	data_[index].title.c_str(), bin);
2364	+	painCave.isFatal = 1;
2365	+	simError();
2366	+	} else {
2367	+	rnemdFile_ << "\t" << s[0] << "\t" << s[1] << "\t" << s[2];
2368	+	}
2369	+	}
2370		}
2371	+

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines