Gromacs
2025dev20240614602a366

#include <gromacs/applied_forces/awh/bias.h>
A bias acting on a multidimensional coordinate.
At each step AWH should provide its biases with updated values of their coordinates. Each bias provides AWH with an updated bias forces and the corresponding potential.
See the user manual for details on the algorithm and equations.
The bias is responsible for keeping and updating a free energy estimate along the coordinate. The bias potential is basically a function of the free energy estimate and so also changes by the update. The free energy update is based on information from coordinate samples collected at a constant bias potential, between updates.
The bias keeps a grid with coordinate points that organizes spatial information about the coordinate. The grid has the same geometry as the coordinate, i.e. they have the same dimensionality and periodicity (if any). The number of points in the grid sets the resolution of the collected data and its extent defines the sampling region of interest.
Each coordinate point has further statistical properties and function values which a grid point does not know about. E.g., for the bias each coordinate point is associated with values of the bias, free energy and target distribution, accumulated sampling weight, etc. For this the bias attaches to each grid point a state. The grid + vector of point states are the bias coordinate points.
The bias has a fairly complex global state keeping track of where the system (coordinate) currently is (CoordState), where it has sampled since the last update (BiasState) and controlling the free energy convergence rate (HistogramSize).
Partly, the complexity comes from the bias having two convergence stages: an initial stage which in an heuristic, nondeterministic way restricts the early convergence rate for sake of robustness; and a final stage where the convergence rate is constant. The length of the initial stage depends on the sampling and is unknown beforehand.
Another complexity comes from the fact that coordinate points, for sake of efficiency in the case of many grid points, are typically only accessed in recently sampled regions even though the free energy update is inherently global and affects all points. The bias allows points thay are nonlocal at the time the update was issued to postpone ("skip", as it is called in the code) the update. A nonlocal point is defined as a point which has not been sampled since the last update. Local points are points that have been sampled since the last update. The (current) set of local points are kept track of by the bias state and reset after every update. An update is called local if it only updates local points. Nonlocal points will temporarily "skip" the update until next time they are local (or when a global update is issued). For this to work, the bias keeps a global "clock" (in HistogramSize) of the number of issued updates. Each PointState also has its own local "clock" with the counting the number of updates it has pulled through. When a point updates its state it asserts that its local clock is synchronized with the global clock.
Public Types  
enum  ThisRankWillDoIO { ThisRankWillDoIO::No, ThisRankWillDoIO::Yes } 
Enum for requesting Bias set up with(out) I/O on this rank. More...  
Public Member Functions  
Bias (int biasIndexInCollection, const AwhParams &awhParams, const AwhBiasParams &awhBiasParams, ArrayRef< const DimParams > dimParams, double beta, double mdTimeStep, const BiasSharing *biasSharing, const std::string &biasInitFilename, ThisRankWillDoIO thisRankWillDoIO, BiasParams::DisableUpdateSkips disableUpdateSkips=BiasParams::DisableUpdateSkips::no)  
Constructor. More...  
void  printInitializationToLog (FILE *fplog) const 
Print information about initialization to log file. More...  
gmx::ArrayRef< const double >  calcForceAndUpdateBias (const awh_dvec coordValue, ArrayRef< const double > neighborLambdaEnergies, ArrayRef< const double > neighborLambdaDhdl, double *awhPotential, double *potentialJump, double t, int64_t step, int64_t seed, FILE *fplog) 
Evolves the bias at every step. More...  
double  calcConvolvedBias (const awh_dvec &coordValue) const 
Calculates the convolved bias for a given coordinate value. More...  
void  restoreStateFromHistory (const AwhBiasHistory *biasHistory, const t_commrec *cr) 
Restore the bias state from history on the main rank and broadcast it. More...  
void  initHistoryFromState (AwhBiasHistory *biasHistory) const 
Allocate and initialize a bias history with the given bias state. More...  
void  updateHistory (AwhBiasHistory *biasHistory) const 
Update the bias history with the current state. More...  
void  doSkippedUpdatesForAllPoints () 
Do all previously skipped updates. Public for use by tests.  
int  ndim () const 
Returns the dimensionality of the bias.  
ArrayRef< const DimParams >  dimParams () const 
Returns the dimension parameters.  
const BiasParams &  params () const 
Returns the bias parameters.  
const BiasState &  state () const 
Returns the global state of the bias.  
int  biasIndex () const 
Returns the index of the bias.  
const awh_dvec &  getGridCoordValue (size_t gridPointIndex) const 
Return the coordinate value for a grid point. More...  
void  updateBiasStateSharedCorrelationTensorTimeIntegral () 
Update the correlation tensor time integral shared across multiple AWH walkers.  
const CorrelationGrid &  forceCorrelationGrid () const 
Return a const reference to the force correlation grid.  
int  numEnergySubblocksToWrite () const 
Return the number of data blocks that have been prepared for writing.  
int  writeToEnergySubblocks (t_enxsubblock *subblock) const 
Write bias data blocks to energy subblocks. More...  
bool  hasFepLambdaDimension () const 
Returns true if the bias has a free energy lambda state dimension at all.  
bool  isSampleCoordStep (int64_t step) const 
Returns whether we should sample the coordinate. More...  

strong 
Enum for requesting Bias set up with(out) I/O on this rank.
Enumerator  

No 
This rank will not do I/O. 
Yes 
This rank will do I/O. 
gmx::Bias::Bias  (  int  biasIndexInCollection, 
const AwhParams &  awhParams,  
const AwhBiasParams &  awhBiasParams,  
ArrayRef< const DimParams >  dimParams,  
double  beta,  
double  mdTimeStep,  
const BiasSharing *  biasSharing,  
const std::string &  biasInitFilename,  
ThisRankWillDoIO  thisRankWillDoIO,  
BiasParams::DisableUpdateSkips  disableUpdateSkips = BiasParams::DisableUpdateSkips::no 

) 
Constructor.
[in]  biasIndexInCollection  Index of the bias in collection. 
[in]  awhParams  AWH parameters. 
[in]  awhBiasParams  Bias parameters. 
[in]  dimParams  Bias dimension parameters. 
[in]  beta  1/(k_B T). 
[in]  mdTimeStep  The MD time step. 
[in]  biasSharing  Pointer to object for sharing bias over simulations, can be nullptr 
[in]  biasInitFilename  Name of file to read PMF and target from. 
[in]  thisRankWillDoIO  Tells whether this MPI rank will do I/O (checkpointing, AWH output), normally (only) the main rank does I/O. 
[in]  disableUpdateSkips  If to disable update skips, useful for testing. 

inline 
Calculates the convolved bias for a given coordinate value.
The convolved bias is the effective bias acting on the coordinate. Since the bias here has arbitrary normalization, this only makes sense as a relative, to other coordinate values, measure of the bias.
[in]  coordValue  The coordinate value. 
gmx::ArrayRef< const double > gmx::Bias::calcForceAndUpdateBias  (  const awh_dvec  coordValue, 
ArrayRef< const double >  neighborLambdaEnergies,  
ArrayRef< const double >  neighborLambdaDhdl,  
double *  awhPotential,  
double *  potentialJump,  
double  t,  
int64_t  step,  
int64_t  seed,  
FILE *  fplog  
) 
Evolves the bias at every step.
At each step the bias step needs to:
[in]  coordValue  The current coordinate value(s). 
[in]  neighborLambdaEnergies  An array containing the energy of the system in neighboring lambdas. The array is of length numLambdas+1, where numLambdas is the number of free energy lambda states. Element 0 in the array is the energy of the current state and elements 1..numLambdas contain the energy of the system in the neighboring lambda states (also including the current state). When there are no free energy lambda state dimensions this can be empty. 
[in]  neighborLambdaDhdl  An array containing the dHdL at the neighboring lambda points. The array is of length numLambdas+1, where numLambdas is the number of free energy lambda states. Element 0 in the array is the dHdL of the current state and elements 1..numLambdas contain the dHdL of the system in the neighboring lambda states (also including the current state). When there are no free energy lambda state dimensions this can be empty. 
[out]  awhPotential  Bias potential. 
[out]  potentialJump  Change in bias potential for this bias. 
[in]  t  Time. 
[in]  step  Time step. 
[in]  seed  Random seed. 
[in,out]  fplog  Log file. 

inline 
Return the coordinate value for a grid point.
[in]  gridPointIndex  The index of the grid point. 
void gmx::Bias::initHistoryFromState  (  AwhBiasHistory *  biasHistory  )  const 
Allocate and initialize a bias history with the given bias state.
This function will be called at the start of a new simulation. Note that only constant data will be initialized here. History data is set by updateHistory.
[in,out]  biasHistory  AWH history to initialize. 
bool gmx::Bias::isSampleCoordStep  (  int64_t  step  )  const 
Returns whether we should sample the coordinate.
[in]  step  The MD step number. 
void gmx::Bias::printInitializationToLog  (  FILE *  fplog  )  const 
Print information about initialization to log file.
Prints information about AWH variables that are set internally but might be of interest to the user.
[in,out]  fplog  Log file, can be nullptr. 
void gmx::Bias::restoreStateFromHistory  (  const AwhBiasHistory *  biasHistory, 
const t_commrec *  cr  
) 
Restore the bias state from history on the main rank and broadcast it.
[in]  biasHistory  Bias history struct, only allowed to be nullptr on worker ranks. 
[in]  cr  The communication record. 
void gmx::Bias::updateHistory  (  AwhBiasHistory *  biasHistory  )  const 
int gmx::Bias::writeToEnergySubblocks  (  t_enxsubblock *  subblock  )  const 
Write bias data blocks to energy subblocks.
[in,out]  subblock  Energy subblocks to write to. 