2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2015,2016,2017,2018, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
38 #include "biaswriter.h"
44 #include "gromacs/awh/awh.h"
45 #include "gromacs/mdtypes/awh-params.h"
46 #include "gromacs/mdtypes/commrec.h"
47 #include "gromacs/trajectory/energyframe.h"
48 #include "gromacs/utility/gmxassert.h"
49 #include "gromacs/utility/smalloc.h"
52 #include "correlationgrid.h"
54 #include "pointstate.h"
63 * Map the output entry type to a normalization type.
65 * The data is written to energy file blocks in the order given by
66 * the iterator of this map, which is based on the enum value
67 * (and matches the order of the lines below).
69 const std::map<AwhOutputEntryType, Normalization> outputTypeToNormalization =
71 { AwhOutputEntryType::MetaData, Normalization::None },
72 { AwhOutputEntryType::CoordValue, Normalization::Coordinate },
73 { AwhOutputEntryType::Pmf, Normalization::FreeEnergy },
74 { AwhOutputEntryType::Bias, Normalization::FreeEnergy },
75 { AwhOutputEntryType::Visits, Normalization::Distribution },
76 { AwhOutputEntryType::Weights, Normalization::Distribution },
77 { AwhOutputEntryType::Target, Normalization::Distribution },
78 { AwhOutputEntryType::ForceCorrelationVolume, Normalization::Distribution },
79 { AwhOutputEntryType::FrictionTensor, Normalization::None }
83 * Gets the coordinate normalization value for the given dimension.
85 * \param[in] bias The AWH bias.
86 * \param[in] dimIndex Dimensional index.
87 * \returns the coordinate normalization value.
89 float getCoordNormalizationValue(const Bias &bias,
92 /* AWH may use different units internally but here we convert to user units */
93 return bias.dimParams()[dimIndex].scaleInternalToUserInput(1);
97 * Gets the normalization value for the given output entry type.
99 * \param[in] outputType Output entry type.
100 * \param[in] bias The AWH bias.
101 * \param[in] numBlocks The number of blocks for this output type.
102 * \returns the normalization value.
104 float getNormalizationValue(AwhOutputEntryType outputType,
108 float normalizationValue = 0;
112 case AwhOutputEntryType::CoordValue:
113 normalizationValue = getCoordNormalizationValue(bias, numBlocks);
115 case AwhOutputEntryType::Visits:
116 case AwhOutputEntryType::Weights:
117 case AwhOutputEntryType::Target:
118 normalizationValue = static_cast<float>(bias.state().points().size());
120 case AwhOutputEntryType::ForceCorrelationVolume:
121 normalizationValue = static_cast<double>(bias.state().points().size());
127 return normalizationValue;
132 AwhEnergyBlock::AwhEnergyBlock(int numPoints,
133 Normalization normalizationType,
134 float normalizationValue) :
135 normalizationType(normalizationType),
136 normalizationValue(normalizationValue),
141 BiasWriter::BiasWriter(const Bias &bias)
143 std::map<AwhOutputEntryType, int> outputTypeNumBlock; /* Number of blocks per output type */
145 /* Different output variable types need different number of blocks.
146 * We keep track of the starting block for each variable.
149 for (const auto &pair : outputTypeToNormalization)
151 const AwhOutputEntryType outputType = pair.first;
153 outputTypeToBlock_[outputType] = blockCount;
155 if (outputType == AwhOutputEntryType::CoordValue)
157 outputTypeNumBlock[outputType] = bias.ndim();
159 else if (outputType == AwhOutputEntryType::FrictionTensor)
161 outputTypeNumBlock[outputType] = bias.forceCorrelationGrid().tensorSize();
165 /* Most output variable types need one block */
166 outputTypeNumBlock[outputType] = 1;
169 blockCount += outputTypeNumBlock[outputType];
172 /* Initialize the data blocks for each variable */
173 for (const auto &pair : outputTypeToNormalization)
175 const AwhOutputEntryType outputType = pair.first;
177 if (outputType == AwhOutputEntryType::MetaData)
179 numPoints = static_cast<int>(AwhOutputMetaData::Count);
183 numPoints = bias.state().points().size();
185 for (int b = 0; b < outputTypeNumBlock[outputType]; b++)
187 block_.emplace_back(numPoints,
189 getNormalizationValue(outputType, bias, b));
195 * Normalizes block data for output.
197 * \param[in,out] block The block to normalize.
198 * \param[in] bias The AWH bias.
200 static void normalizeBlock(AwhEnergyBlock *block, const Bias &bias)
202 gmx::ArrayRef<float> data = block->data();
204 /* Here we operate on float data (which is accurate enough, since it
205 * is statistical data that will never reach full float precision).
206 * But since we can have very many data points, we sum into a double.
209 float minValue = GMX_FLOAT_MAX;
212 switch (block->normalizationType)
214 case Normalization::None:
216 case Normalization::Coordinate:
217 /* Normalize coordinate values by a scale factor */
218 for (float &point : data)
220 point *= block->normalizationValue;
223 case Normalization::FreeEnergy:
224 /* Normalize free energy values by subtracting the minimum value */
225 for (gmx::index index = 0; index < data.size(); index++)
227 if (bias.state().points()[index].inTargetRegion() && data[index] < minValue)
229 minValue = data[index];
232 for (gmx::index index = 0; index < data.size(); index++)
234 if (bias.state().points()[index].inTargetRegion())
236 data[index] -= minValue;
241 case Normalization::Distribution:
242 /* Normalize distribution values by normalizing their sum */
243 for (float &point : data)
249 recipNorm = block->normalizationValue/static_cast<float>(sum);
251 for (float &point : data)
257 GMX_RELEASE_ASSERT(false, "Unknown AWH normalization type");
262 void BiasWriter::transferMetaDataToWriter(gmx::index metaDataIndex,
263 AwhOutputMetaData metaDataType,
266 gmx::ArrayRef<float> data = block_[getVarStartBlock(AwhOutputEntryType::MetaData)].data();
267 GMX_ASSERT(metaDataIndex < data.size(), "Attempt to transfer AWH meta data to block for index out of range");
269 /* Transfer the point data of this variable to the right block(s) */
270 switch (metaDataType)
272 case AwhOutputMetaData::NumBlock:
273 /* The number of subblocks per awh (needed by gmx_energy) */
274 data[metaDataIndex] = static_cast<double>(block_.size());
275 /* Note: a single subblock takes only a single type and we need doubles. */
277 case AwhOutputMetaData::TargetError:
278 /* The theoretical target error */
279 data[metaDataIndex] = bias.params().initialErrorInKT*std::sqrt(bias.params().initialHistogramSize/bias.state().histogramSize().histogramSize());
281 case AwhOutputMetaData::ScaledSampleWeight:
282 /* The logarithm of the sample weight relative to a sample weight of 1 at the initial time.
283 In the normal case: this will increase in the initial stage and then stay at a constant value. */
284 data[metaDataIndex] = bias.state().histogramSize().logScaledSampleWeight();
286 case AwhOutputMetaData::Count:
292 BiasWriter::transferPointDataToWriter(AwhOutputEntryType outputType,
295 gmx::ArrayRef<const float> pmf)
297 /* The starting block index of this output type.
298 * Note that some variables need several (contiguous) blocks.
300 int blockStart = getVarStartBlock(outputType);
301 GMX_ASSERT(pointIndex < static_cast<int>(block_[blockStart].data().size()), "Attempt to transfer AWH data to block for point index out of range");
303 const CorrelationGrid &forceCorrelation = bias.forceCorrelationGrid();
304 int numCorrelation = forceCorrelation.tensorSize();
306 /* Transfer the point data of this variable to the right block(s) */
310 case AwhOutputEntryType::MetaData:
311 GMX_RELEASE_ASSERT(false, "MetaData is handled by a different function");
313 case AwhOutputEntryType::CoordValue:
315 const awh_dvec &coordValue = bias.getGridCoordValue(pointIndex);
316 for (int d = 0; d < bias.ndim(); d++)
318 block_[b].data()[pointIndex] = coordValue[d];
323 case AwhOutputEntryType::Pmf:
324 block_[b].data()[pointIndex] = bias.state().points()[pointIndex].inTargetRegion() ? pmf[pointIndex] : 0;
326 case AwhOutputEntryType::Bias:
328 const awh_dvec &coordValue = bias.getGridCoordValue(pointIndex);
329 block_[b].data()[pointIndex] = bias.state().points()[pointIndex].inTargetRegion() ? bias.calcConvolvedBias(coordValue) : 0;
332 case AwhOutputEntryType::Visits:
333 block_[b].data()[pointIndex] = bias.state().points()[pointIndex].numVisitsTot();
335 case AwhOutputEntryType::Weights:
336 block_[b].data()[pointIndex] = bias.state().points()[pointIndex].weightSumTot();
338 case AwhOutputEntryType::Target:
339 block_[b].data()[pointIndex] = bias.state().points()[pointIndex].target();
341 case AwhOutputEntryType::ForceCorrelationVolume:
342 block_[b].data()[pointIndex] = forceCorrelation.tensors()[pointIndex].getVolumeElement(forceCorrelation.dtSample);
344 case AwhOutputEntryType::FrictionTensor:
345 /* Store force correlation in units of friction, i.e. time/length^2 */
346 for (int n = 0; n < numCorrelation; n++)
348 block_[b].data()[pointIndex] = forceCorrelation.tensors()[pointIndex].getTimeIntegral(n, forceCorrelation.dtSample);
353 GMX_RELEASE_ASSERT(false, "Unknown AWH output variable");
358 void BiasWriter::prepareBiasOutput(const Bias &bias)
360 /* Pack the AWH data into the writer data. */
362 /* Evaluate the PMF for all points */
363 gmx::ArrayRef<float> pmf = block_[getVarStartBlock(AwhOutputEntryType::Pmf)].data();
364 bias.state().getPmf(pmf);
366 /* Pack the the data point by point.
367 * Unfortunately we can not loop over a class enum, so we cast to int.
368 * \todo Use strings instead of enum when we port the output to TNG.
370 for (int i = 0; i < static_cast<int>(AwhOutputMetaData::Count); i++)
372 transferMetaDataToWriter(i, static_cast<AwhOutputMetaData>(i), bias);
374 for (const auto &pair : outputTypeToNormalization)
376 const AwhOutputEntryType outputType = pair.first;
377 /* Skip metadata (transfered above) and unused blocks */
378 if (outputType == AwhOutputEntryType::MetaData || !hasVarBlock(outputType))
382 for (size_t m = 0; m < bias.state().points().size(); m++)
384 transferPointDataToWriter(outputType, m, bias, pmf);
388 /* For looks of the output, normalize it */
389 for (AwhEnergyBlock &block : block_)
391 normalizeBlock(&block, bias);
395 int BiasWriter::writeToEnergySubblocks(const Bias &bias,
398 prepareBiasOutput(bias);
400 for (size_t b = 0; b < block_.size(); b++)
402 sub[b].type = xdr_datatype_float;
403 sub[b].nr = block_[b].data().size();
404 sub[b].fval = block_[b].data().data();
407 return block_.size();