Page MenuHomec4science
Files F92008203

crml_gpu_memory2.h
No OneTemporary
Actions

File Metadata

Created
Sat, Nov 16, 14:01

crml_gpu_memory2.h
View Options

/* ----------------------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov
Copyright (2003) Sandia Corporation. Under the terms of Contract
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
certain rights in this software. This software is distributed under
the GNU General Public License.
See the README file in the top-level LAMMPS directory.
------------------------------------------------------------------------- */
/* ----------------------------------------------------------------------
Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
------------------------------------------------------------------------- */
#ifndef CRML_GPU_MEMORY2_H
#define CRML_GPU_MEMORY2_H
#include "charge_gpu_memory2.h"
template <class numtyp, class acctyp>
class CRML_GPU_Memory2 : public ChargeGPUMemory2<numtyp, acctyp> {
public:
CRML_GPU_Memory2();
~CRML_GPU_Memory2();
/// Clear any previous data and set up for a new LAMMPS run
/** \param max_nbors initial number of rows in the neighbor matrix
* \param cell_size cutoff + skin
* \param gpu_split fraction of particles handled by device
*
* Returns:
* - 0 if successfull
* - -1 if fix gpu not found
* - -3 if there is an out of memory error
* - -4 if the GPU library was not compiled for GPU
* - -5 Double precision is not supported on card **/
int init(const int ntypes, double host_cut_bothsq,
double **host_lj1, double **host_lj2, double **host_lj3,
double **host_lj4, double **host_offset, double *host_special_lj,
const int nlocal, const int nall, const int max_nbors,
const int maxspecial, const double cell_size,
const double gpu_split, FILE *screen, double host_cut_ljsq,
const double host_cut_coulsq, double *host_special_coul,
const double qqrd2e, const double g_ewald,
const double cut_lj_innersq, const double denom_lj,
double **epsilon, double **sigma, const bool mix_arithmetic);
/// Clear all host and device data
/** \note This is called at the beginning of the init() routine **/
void clear();
/// Returns memory usage on device per atom
int bytes_per_atom(const int max_nbors) const;
/// Total host memory used by library for pair style
double host_memory_usage() const;
// --------------------------- TYPE DATA --------------------------
/// x = lj1, y = lj2, z = lj3, w = lj4
UCL_D_Vec<numtyp4> lj1;
/// x = epsilon, y = sigma
UCL_D_Vec<numtyp2> ljd;
/// Special LJ values [0-3] and Special Coul values [4-7]
UCL_D_Vec<numtyp> sp_lj;
/// If atom type constants fit in shared memory, use fast kernels
bool shared_types;
/// Number of atom types
int _lj_types;
numtyp _qqrd2e, _g_ewald, _denom_lj;
numtyp _cut_coulsq, _cut_bothsq, _cut_ljsq, _cut_lj_innersq;
private:
bool _allocated;
void loop(const bool _eflag, const bool _vflag);
};
#endif

Event Timeline

c4science · Help