PmeAtomComm Class Reference

#include <gromacs/ewald/pme_internal.h>

Collaboration diagram for PmeAtomComm:

Description

Data structure for coordinating transfers between PME ranks along one dimension.

Also used for passing coordinates, coefficients and forces to and from PME routines.

Public Member Functions
	PmeAtomComm (MPI_Comm PmeMpiCommunicator, int numThreads, int pmeOrder, int dimIndex, bool doSpread)
	Constructor, PmeMpiCommunicator is the communicator for this dimension.
void	setNumAtoms (int numAtoms)
	Set the atom count and when necessary resizes atom buffers.
int	numAtoms () const
	Returns the atom count.
gmx::ArrayRef< int >	sendCount ()
	Returns the number of atoms to send to each rank.

Public Attributes
int	dimind = 0
	The index of the dimension, 0=x, 1=y.
int	nslab = 1
	The number of slabs and ranks this dimension is decomposed over.
int	slabIndex = 0
	Our slab index, our MPI rank is equal to this number.
MPI_Comm	mpi_comm
	Communicator for this dimension.
std::vector< SlabCommSetup >	slabCommSetup
	Communication setup for each slab, ordered as alternating forward/backward and increasing slab shift.
int	maxshift = 0
	The maximum communication distance counted in MPI ranks.
std::vector< int >	bufferIndices
	Working buffer indices, indexed with the slab index.
FastVector< int >	pd
	The target slab index for each particle.
std::vector< std::vector< int > >	count_thread
	Target particle counts for each slab, for each thread.
gmx::ArrayRef< const gmx::RVec >	x
	The coordinates.
gmx::ArrayRef< const real >	coefficient
	The coefficient, charges or LJ C6.
gmx::ArrayRef< gmx::RVec >	f
	The forces.
FastVector< gmx::RVec >	xBuffer
	Coordinate buffer, used only with nslab > 1.
FastVector< real >	coefficientBuffer
	Coefficient buffer, used only with nslab > 1.
FastVector< gmx::RVec >	fBuffer
	Force buffer, used only with nslab > 1.
bool	bSpread
	Tells whether these coordinates are used for spreading.
int	pme_order
	The PME order.
FastVector< gmx::IVec >	idx
	The grid index per atom.
FastVector< gmx::RVec >	fractx
	Fractional atom coordinates relative to the lower cell boundary.
int	nthread
	The number of threads to use in PME.
FastVector< int >	thread_idx
	Thread index for each atom.
std::vector< AtomToThreadMap >	threadMap
std::vector< splinedata_t >	spline

---

The documentation for this class was generated from the following file:

• src/gromacs/ewald/pme_internal.h