Homec4science
Diffusion lammps (efficient_neuronet) Browse

lammps/src/KOKKOS/pair_lj_charmm_coul_long_kokkos.cpp
34785af4a6a0efficient_neuronet

Last Change
Raw File

pair_lj_charmm_coul_long_kokkos.cpp
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 author: Ray Shan (SNL)
------------------------------------------------------------------------- */
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "pair_lj_charmm_coul_long_kokkos.h"
#include "kokkos.h"
#include "atom_kokkos.h"
#include "comm.h"
#include "force.h"
#include "neighbor.h"
#include "neigh_list.h"
#include "neigh_request.h"
#include "update.h"
#include "integrate.h"
#include "respa.h"
#include "math_const.h"
#include "memory.h"
#include "error.h"
#include "atom_masks.h"
using namespace LAMMPS_NS;
using namespace MathConst;
#define KOKKOS_CUDA_MAX_THREADS 256
#define KOKKOS_CUDA_MIN_BLOCKS 8
#define EWALD_F 1.12837917
#define EWALD_P 0.3275911
#define A1 0.254829592
#define A2 -0.284496736
#define A3 1.421413741
#define A4 -1.453152027
#define A5 1.061405429
/* ---------------------------------------------------------------------- */
template<class DeviceType>
PairLJCharmmCoulLongKokkos<DeviceType>::PairLJCharmmCoulLongKokkos(LAMMPS *lmp):PairLJCharmmCoulLong(lmp)
{
respa_enable = 0;
atomKK = (AtomKokkos *) atom;
execution_space = ExecutionSpaceFromDevice<DeviceType>::space;
datamask_read = X_MASK | F_MASK | TYPE_MASK | Q_MASK | ENERGY_MASK | VIRIAL_MASK;
datamask_modify = F_MASK | ENERGY_MASK | VIRIAL_MASK;
cutsq = NULL;
cut_ljsq = 0.0;
cut_coulsq = 0.0;
}
/* ---------------------------------------------------------------------- */
template<class DeviceType>
PairLJCharmmCoulLongKokkos<DeviceType>::~PairLJCharmmCoulLongKokkos()
{
if (!copymode) {
memory->destroy_kokkos(k_eatom,eatom);
memory->destroy_kokkos(k_vatom,vatom);
k_cutsq = DAT::tdual_ffloat_2d();
k_cut_ljsq = DAT::tdual_ffloat_2d();
k_cut_coulsq = DAT::tdual_ffloat_2d();
memory->sfree(cutsq);
//memory->sfree(cut_ljsq);
//memory->sfree(cut_coulsq);
eatom = NULL;
vatom = NULL;
cutsq = NULL;
cut_ljsq = 0.0;
cut_coulsq = 0.0;
}
}
/* ---------------------------------------------------------------------- */
template<class DeviceType>
void PairLJCharmmCoulLongKokkos<DeviceType>::cleanup_copy() {
// WHY needed: this prevents parent copy from deallocating any arrays
allocated = 0;
cutsq = NULL;
cut_ljsq = 0.0;
eatom = NULL;
vatom = NULL;
ftable = NULL;
}
/* ---------------------------------------------------------------------- */
template<class DeviceType>
void PairLJCharmmCoulLongKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
{
eflag = eflag_in;
vflag = vflag_in;
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;
atomKK->sync(execution_space,datamask_read);
k_cutsq.template sync<DeviceType>();
k_cut_ljsq.template sync<DeviceType>();
k_cut_coulsq.template sync<DeviceType>();
k_params.template sync<DeviceType>();
if (eflag || vflag) atomKK->modified(execution_space,datamask_modify);
else atomKK->modified(execution_space,F_MASK);
x = atomKK->k_x.view<DeviceType>();
c_x = atomKK->k_x.view<DeviceType>();
f = atomKK->k_f.view<DeviceType>();
q = atomKK->k_q.view<DeviceType>();
type = atomKK->k_type.view<DeviceType>();
nlocal = atom->nlocal;
nall = atom->nlocal + atom->nghost;
special_lj[0] = force->special_lj[0];
special_lj[1] = force->special_lj[1];
special_lj[2] = force->special_lj[2];
special_lj[3] = force->special_lj[3];
special_coul[0] = force->special_coul[0];
special_coul[1] = force->special_coul[1];
special_coul[2] = force->special_coul[2];
special_coul[3] = force->special_coul[3];
qqrd2e = force->qqrd2e;
newton_pair = force->newton_pair;
// loop over neighbors of my atoms
copymode = 1;
EV_FLOAT ev;
if(ncoultablebits)
ev = pair_compute<PairLJCharmmCoulLongKokkos<DeviceType>,CoulLongTable<1> >
(this,(NeighListKokkos<DeviceType>*)list);
else
ev = pair_compute<PairLJCharmmCoulLongKokkos<DeviceType>,CoulLongTable<0> >
(this,(NeighListKokkos<DeviceType>*)list);
DeviceType::fence();
if (eflag) {
eng_vdwl += ev.evdwl;
eng_coul += ev.ecoul;
}
if (vflag_global) {
virial[0] += ev.v[0];
virial[1] += ev.v[1];
virial[2] += ev.v[2];
virial[3] += ev.v[3];
virial[4] += ev.v[4];
virial[5] += ev.v[5];
}
if (vflag_fdotr) pair_virial_fdotr_compute(this);
copymode = 0;
}
/* ----------------------------------------------------------------------
compute LJ CHARMM pair force between atoms i and j
---------------------------------------------------------------------- */
template<class DeviceType>
template<bool STACKPARAMS, class Specialisation>
KOKKOS_INLINE_FUNCTION
F_FLOAT PairLJCharmmCoulLongKokkos<DeviceType>::
compute_fpair(const F_FLOAT& rsq, const int& i, const int&j,
const int& itype, const int& jtype) const {
const F_FLOAT r2inv = 1.0/rsq;
const F_FLOAT r6inv = r2inv*r2inv*r2inv;
F_FLOAT forcelj, switch1, switch2, englj;
forcelj = r6inv *
((STACKPARAMS?m_params[itype][jtype].lj1:params(itype,jtype).lj1)*r6inv -
(STACKPARAMS?m_params[itype][jtype].lj2:params(itype,jtype).lj2));
if (rsq > cut_lj_innersq) {
switch1 = (cut_ljsq-rsq) * (cut_ljsq-rsq) *
(cut_ljsq + 2.0*rsq - 3.0*cut_lj_innersq) / denom_lj;
switch2 = 12.0*rsq * (cut_ljsq-rsq) * (rsq-cut_lj_innersq) / denom_lj;
englj = r6inv *
((STACKPARAMS?m_params[itype][jtype].lj3:params(itype,jtype).lj3)*r6inv -
(STACKPARAMS?m_params[itype][jtype].lj4:params(itype,jtype).lj4));
forcelj = forcelj*switch1 + englj*switch2;
}
return forcelj*r2inv;
}
/* ----------------------------------------------------------------------
compute LJ CHARMM pair potential energy between atoms i and j
---------------------------------------------------------------------- */
template<class DeviceType>
template<bool STACKPARAMS, class Specialisation>
KOKKOS_INLINE_FUNCTION
F_FLOAT PairLJCharmmCoulLongKokkos<DeviceType>::
compute_evdwl(const F_FLOAT& rsq, const int& i, const int&j,
const int& itype, const int& jtype) const {
const F_FLOAT r2inv = 1.0/rsq;
const F_FLOAT r6inv = r2inv*r2inv*r2inv;
F_FLOAT englj, switch1;
englj = r6inv *
((STACKPARAMS?m_params[itype][jtype].lj3:params(itype,jtype).lj3)*r6inv -
(STACKPARAMS?m_params[itype][jtype].lj4:params(itype,jtype).lj4));
if (rsq > cut_lj_innersq) {
switch1 = (cut_ljsq-rsq) * (cut_ljsq-rsq) *
(cut_ljsq + 2.0*rsq - 3.0*cut_lj_innersq) / denom_lj;
englj *= switch1;
}
return englj;
}
/* ----------------------------------------------------------------------
compute coulomb pair force between atoms i and j
---------------------------------------------------------------------- */
template<class DeviceType>
template<bool STACKPARAMS, class Specialisation>
KOKKOS_INLINE_FUNCTION
F_FLOAT PairLJCharmmCoulLongKokkos<DeviceType>::
compute_fcoul(const F_FLOAT& rsq, const int& i, const int&j,
const int& itype, const int& jtype, const F_FLOAT& factor_coul, const F_FLOAT& qtmp) const {
if(Specialisation::DoTable && rsq > tabinnersq) {
union_int_float_t rsq_lookup;
rsq_lookup.f = rsq;
const int itable = (rsq_lookup.i & ncoulmask) >> ncoulshiftbits;
const F_FLOAT fraction = (rsq_lookup.f - d_rtable[itable]) * d_drtable[itable];
const F_FLOAT table = d_ftable[itable] + fraction*d_dftable[itable];
F_FLOAT forcecoul = qtmp*q[j] * table;
if (factor_coul < 1.0) {
const F_FLOAT table = d_ctable[itable] + fraction*d_dctable[itable];
const F_FLOAT prefactor = qtmp*q[j] * table;
forcecoul -= (1.0-factor_coul)*prefactor;
}
return forcecoul/rsq;
} else {
const F_FLOAT r = sqrt(rsq);
const F_FLOAT grij = g_ewald * r;
const F_FLOAT expm2 = exp(-grij*grij);
const F_FLOAT t = 1.0 / (1.0 + EWALD_P*grij);
const F_FLOAT rinv = 1.0/r;
const F_FLOAT erfc = t * (A1+t*(A2+t*(A3+t*(A4+t*A5)))) * expm2;
const F_FLOAT prefactor = qqrd2e * qtmp*q[j]*rinv;
F_FLOAT forcecoul = prefactor * (erfc + EWALD_F*grij*expm2);
if (factor_coul < 1.0) forcecoul -= (1.0-factor_coul)*prefactor;
return forcecoul*rinv*rinv;
}
}
/* ----------------------------------------------------------------------
compute coulomb pair potential energy between atoms i and j
---------------------------------------------------------------------- */
template<class DeviceType>
template<bool STACKPARAMS, class Specialisation>
KOKKOS_INLINE_FUNCTION
F_FLOAT PairLJCharmmCoulLongKokkos<DeviceType>::
compute_ecoul(const F_FLOAT& rsq, const int& i, const int&j,
const int& itype, const int& jtype, const F_FLOAT& factor_coul, const F_FLOAT& qtmp) const {
if(Specialisation::DoTable && rsq > tabinnersq) {
union_int_float_t rsq_lookup;
rsq_lookup.f = rsq;
const int itable = (rsq_lookup.i & ncoulmask) >> ncoulshiftbits;
const F_FLOAT fraction = (rsq_lookup.f - d_rtable[itable]) * d_drtable[itable];
const F_FLOAT table = d_etable[itable] + fraction*d_detable[itable];
F_FLOAT ecoul = qtmp*q[j] * table;
if (factor_coul < 1.0) {
const F_FLOAT table = d_ctable[itable] + fraction*d_dctable[itable];
const F_FLOAT prefactor = qtmp*q[j] * table;
ecoul -= (1.0-factor_coul)*prefactor;
}
return ecoul;
} else {
const F_FLOAT r = sqrt(rsq);
const F_FLOAT grij = g_ewald * r;
const F_FLOAT expm2 = exp(-grij*grij);
const F_FLOAT t = 1.0 / (1.0 + EWALD_P*grij);
const F_FLOAT erfc = t * (A1+t*(A2+t*(A3+t*(A4+t*A5)))) * expm2;
const F_FLOAT prefactor = qqrd2e * qtmp*q[j]/r;
F_FLOAT ecoul = prefactor * erfc;
if (factor_coul < 1.0) ecoul -= (1.0-factor_coul)*prefactor;
return ecoul;
}
}
/* ----------------------------------------------------------------------
allocate all arrays
------------------------------------------------------------------------- */
template<class DeviceType>
void PairLJCharmmCoulLongKokkos<DeviceType>::allocate()
{
PairLJCharmmCoulLong::allocate();
int n = atom->ntypes;
memory->destroy(cutsq);
memory->create_kokkos(k_cutsq,cutsq,n+1,n+1,"pair:cutsq");
d_cutsq = k_cutsq.template view<DeviceType>();
//memory->destroy(cut_ljsq);
memory->create_kokkos(k_cut_ljsq,n+1,n+1,"pair:cut_ljsq");
d_cut_ljsq = k_cut_ljsq.template view<DeviceType>();
memory->create_kokkos(k_cut_coulsq,n+1,n+1,"pair:cut_coulsq");
d_cut_coulsq = k_cut_coulsq.template view<DeviceType>();
k_params = Kokkos::DualView<params_lj_coul**,Kokkos::LayoutRight,DeviceType>("PairLJCharmmCoulLong::params",n+1,n+1);
params = k_params.d_view;
}
template<class DeviceType>
void PairLJCharmmCoulLongKokkos<DeviceType>::init_tables(double cut_coul, double *cut_respa)
{
Pair::init_tables(cut_coul,cut_respa);
typedef typename ArrayTypes<DeviceType>::t_ffloat_1d table_type;
typedef typename ArrayTypes<LMPHostType>::t_ffloat_1d host_table_type;
int ntable = 1;
for (int i = 0; i < ncoultablebits; i++) ntable *= 2;
// Copy rtable and drtable
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
for(int i = 0; i < ntable; i++) {
h_table(i) = rtable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_rtable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
for(int i = 0; i < ntable; i++) {
h_table(i) = drtable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_drtable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
// Copy ftable and dftable
for(int i = 0; i < ntable; i++) {
h_table(i) = ftable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_ftable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
for(int i = 0; i < ntable; i++) {
h_table(i) = dftable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_dftable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
// Copy ctable and dctable
for(int i = 0; i < ntable; i++) {
h_table(i) = ctable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_ctable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
for(int i = 0; i < ntable; i++) {
h_table(i) = dctable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_dctable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
// Copy etable and detable
for(int i = 0; i < ntable; i++) {
h_table(i) = etable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_etable = d_table;
}
{
host_table_type h_table("HostTable",ntable);
table_type d_table("DeviceTable",ntable);
for(int i = 0; i < ntable; i++) {
h_table(i) = detable[i];
}
Kokkos::deep_copy(d_table,h_table);
d_detable = d_table;
}
}
/* ----------------------------------------------------------------------
init specific to this pair style
------------------------------------------------------------------------- */
template<class DeviceType>
void PairLJCharmmCoulLongKokkos<DeviceType>::init_style()
{
PairLJCharmmCoulLong::init_style();
// error if rRESPA with inner levels
if (update->whichflag == 1 && strstr(update->integrate_style,"respa")) {
int respa = 0;
if (((Respa *) update->integrate)->level_inner >= 0) respa = 1;
if (((Respa *) update->integrate)->level_middle >= 0) respa = 2;
if (respa)
error->all(FLERR,"Cannot use Kokkos pair style with rRESPA inner/middle");
}
// irequest = neigh request made by parent class
neighflag = lmp->kokkos->neighflag;
int irequest = neighbor->nrequest - 1;
neighbor->requests[irequest]->
kokkos_host = Kokkos::Impl::is_same<DeviceType,LMPHostType>::value &&
!Kokkos::Impl::is_same<DeviceType,LMPDeviceType>::value;
neighbor->requests[irequest]->
kokkos_device = Kokkos::Impl::is_same<DeviceType,LMPDeviceType>::value;
if (neighflag == FULL) {
neighbor->requests[irequest]->full = 1;
neighbor->requests[irequest]->half = 0;
neighbor->requests[irequest]->full_cluster = 0;
} else if (neighflag == HALF || neighflag == HALFTHREAD) {
neighbor->requests[irequest]->full = 0;
neighbor->requests[irequest]->half = 1;
neighbor->requests[irequest]->full_cluster = 0;
} else {
error->all(FLERR,"Cannot use chosen neighbor list style with lj/charmm/coul/long/kk");
}
}
/* ----------------------------------------------------------------------
init for one type pair i,j and corresponding j,i
------------------------------------------------------------------------- */
template<class DeviceType>
double PairLJCharmmCoulLongKokkos<DeviceType>::init_one(int i, int j)
{
double cutone = PairLJCharmmCoulLong::init_one(i,j);
double cut_ljsqm = cut_ljsq;
double cut_coulsqm = cut_coulsq;
k_params.h_view(i,j).lj1 = lj1[i][j];
k_params.h_view(i,j).lj2 = lj2[i][j];
k_params.h_view(i,j).lj3 = lj3[i][j];
k_params.h_view(i,j).lj4 = lj4[i][j];
//k_params.h_view(i,j).offset = offset[i][j];
k_params.h_view(i,j).cut_ljsq = cut_ljsqm;
k_params.h_view(i,j).cut_coulsq = cut_coulsqm;
k_params.h_view(j,i) = k_params.h_view(i,j);
if(i<MAX_TYPES_STACKPARAMS+1 && j<MAX_TYPES_STACKPARAMS+1) {
m_params[i][j] = m_params[j][i] = k_params.h_view(i,j);
m_cutsq[j][i] = m_cutsq[i][j] = cutone*cutone;
m_cut_ljsq[j][i] = m_cut_ljsq[i][j] = cut_ljsqm;
m_cut_coulsq[j][i] = m_cut_coulsq[i][j] = cut_coulsqm;
}
k_cutsq.h_view(i,j) = cutone*cutone;
k_cutsq.h_view(j,i) = k_cutsq.h_view(i,j);
k_cutsq.template modify<LMPHostType>();
k_cut_ljsq.h_view(i,j) = cut_ljsqm;
k_cut_ljsq.h_view(j,i) = k_cut_ljsq.h_view(i,j);
k_cut_ljsq.template modify<LMPHostType>();
k_cut_coulsq.h_view(i,j) = cut_coulsqm;
k_cut_coulsq.h_view(j,i) = k_cut_coulsq.h_view(i,j);
k_cut_coulsq.template modify<LMPHostType>();
k_params.template modify<LMPHostType>();
return cutone;
}
template class PairLJCharmmCoulLongKokkos<LMPDeviceType>;
#ifdef KOKKOS_HAVE_CUDA
template class PairLJCharmmCoulLongKokkos<LMPHostType>;
#endif
c4science · Help