496 lines
16 KiB
C++
496 lines
16 KiB
C++
// clang-format off
|
|
/* -*- c++ -*- ----------------------------------------------------------
|
|
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
|
https://www.lammps.org/, 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: Rodrigo Canales (RWTH Aachen University)
|
|
------------------------------------------------------------------------- */
|
|
|
|
#include "pair_buck_intel.h"
|
|
|
|
#include "atom.h"
|
|
#include "comm.h"
|
|
#include "error.h"
|
|
#include "force.h"
|
|
#include "math_const.h"
|
|
#include "memory.h"
|
|
#include "modify.h"
|
|
#include "neigh_list.h"
|
|
#include "neigh_request.h"
|
|
#include "neighbor.h"
|
|
#include "suffix.h"
|
|
|
|
#include <cmath>
|
|
#include <cstring>
|
|
|
|
using namespace LAMMPS_NS;
|
|
using namespace MathConst;
|
|
|
|
#define C_FORCE_T typename ForceConst<flt_t>::c_force_t
|
|
#define C_ENERGY_T typename ForceConst<flt_t>::c_energy_t
|
|
|
|
PairBuckIntel::PairBuckIntel(LAMMPS *lmp) : PairBuck(lmp)
|
|
{
|
|
suffix_flag |= Suffix::INTEL;
|
|
}
|
|
|
|
void PairBuckIntel::compute(int eflag, int vflag)
|
|
{
|
|
if (fix->precision()==FixIntel::PREC_MODE_MIXED)
|
|
compute<float,double>(eflag, vflag, fix->get_mixed_buffers(),
|
|
force_const_single);
|
|
else if (fix->precision()==FixIntel::PREC_MODE_DOUBLE)
|
|
compute<double,double>(eflag, vflag, fix->get_double_buffers(),
|
|
force_const_double);
|
|
else
|
|
compute<float,float>(eflag, vflag, fix->get_single_buffers(),
|
|
force_const_single);
|
|
|
|
fix->balance_stamp();
|
|
vflag_fdotr = 0;
|
|
}
|
|
|
|
template <class flt_t, class acc_t>
|
|
void PairBuckIntel::compute(int eflag, int vflag,
|
|
IntelBuffers<flt_t,acc_t> *buffers,
|
|
const ForceConst<flt_t> &fc)
|
|
{
|
|
ev_init(eflag,vflag);
|
|
if (vflag_atom)
|
|
error->all(FLERR,"INTEL package does not support per-atom stress");
|
|
if (vflag && !vflag_fdotr && force->newton_pair)
|
|
error->all(FLERR,"INTEL package does not support pair_modify nofdotr "
|
|
"with newton on");
|
|
|
|
const int inum = list->inum;
|
|
const int nthreads = comm->nthreads;
|
|
const int host_start = fix->host_start_pair();
|
|
const int offload_end = fix->offload_end_pair();
|
|
const int ago = neighbor->ago;
|
|
|
|
if (ago != 0 && fix->separate_buffers() == 0) {
|
|
fix->start_watch(TIME_PACK);
|
|
|
|
int packthreads;
|
|
if (nthreads > INTEL_HTHREADS) packthreads = nthreads;
|
|
else packthreads = 1;
|
|
#if defined(_OPENMP)
|
|
#pragma omp parallel if (packthreads > 1)
|
|
#endif
|
|
{
|
|
int ifrom, ito, tid;
|
|
IP_PRE_omp_range_id_align(ifrom, ito, tid, atom->nlocal + atom->nghost,
|
|
packthreads, sizeof(ATOM_T));
|
|
buffers->thr_pack(ifrom,ito,ago);
|
|
}
|
|
fix->stop_watch(TIME_PACK);
|
|
}
|
|
|
|
int ovflag = 0;
|
|
if (vflag_fdotr) ovflag = 2;
|
|
else if (vflag) ovflag = 1;
|
|
if (eflag) {
|
|
if (force->newton_pair) {
|
|
eval<1,1>(1, ovflag, buffers, fc, 0, offload_end);
|
|
eval<1,1>(0, ovflag, buffers, fc, host_start, inum);
|
|
} else {
|
|
eval<1,0>(1, ovflag, buffers, fc, 0, offload_end);
|
|
eval<1,0>(0, ovflag, buffers, fc, host_start, inum);
|
|
}
|
|
} else {
|
|
if (force->newton_pair) {
|
|
eval<0,1>(1, ovflag, buffers, fc, 0, offload_end);
|
|
eval<0,1>(0, ovflag, buffers, fc, host_start, inum);
|
|
} else {
|
|
eval<0,0>(1, ovflag, buffers, fc, 0, offload_end);
|
|
eval<0,0>(0, ovflag, buffers, fc, host_start, inum);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* ---------------------------------------------------------------------- */
|
|
|
|
template <int EFLAG, int NEWTON_PAIR, class flt_t, class acc_t>
|
|
void PairBuckIntel::eval(const int offload, const int vflag,
|
|
IntelBuffers<flt_t,acc_t> *buffers,
|
|
const ForceConst<flt_t> &fc,
|
|
const int astart, const int aend)
|
|
{
|
|
const int inum = aend - astart;
|
|
if (inum == 0) return;
|
|
int nlocal, nall, minlocal;
|
|
fix->get_buffern(offload, nlocal, nall, minlocal);
|
|
const int ago = neighbor->ago;
|
|
IP_PRE_pack_separate_buffers(fix, buffers, ago, offload, nlocal, nall);
|
|
|
|
ATOM_T * _noalias const x = buffers->get_x(offload);
|
|
|
|
const int * _noalias const ilist = list->ilist;
|
|
const int * _noalias const numneigh = list->numneigh;
|
|
const int ** _noalias const firstneigh = (const int **)list->firstneigh; // NOLINT
|
|
|
|
const flt_t * _noalias const special_lj = fc.special_lj;
|
|
const C_FORCE_T * _noalias const c_force = fc.c_force[0];
|
|
const C_ENERGY_T * _noalias const c_energy = fc.c_energy[0];
|
|
|
|
const int ntypes = atom->ntypes + 1;
|
|
const int eatom = this->eflag_atom;
|
|
|
|
// Determine how much data to transfer
|
|
int x_size, q_size, f_stride, ev_size, separate_flag;
|
|
IP_PRE_get_transfern(ago, NEWTON_PAIR, EFLAG, vflag,
|
|
buffers, offload, fix, separate_flag,
|
|
x_size, q_size, ev_size, f_stride);
|
|
|
|
int tc;
|
|
FORCE_T * _noalias f_start;
|
|
acc_t * _noalias ev_global;
|
|
IP_PRE_get_buffers(offload, buffers, fix, tc, f_start, ev_global);
|
|
|
|
const int nthreads = tc;
|
|
#ifdef _LMP_INTEL_OFFLOAD
|
|
int *overflow = fix->get_off_overflow_flag();
|
|
double *timer_compute = fix->off_watch_pair();
|
|
// Redeclare as local variables for offload
|
|
|
|
if (offload) fix->start_watch(TIME_OFFLOAD_LATENCY);
|
|
#pragma offload target(mic:_cop) if (offload) \
|
|
in(special_lj:length(0) alloc_if(0) free_if(0)) \
|
|
in(c_force, c_energy:length(0) alloc_if(0) free_if(0)) \
|
|
in(firstneigh:length(0) alloc_if(0) free_if(0)) \
|
|
in(numneigh:length(0) alloc_if(0) free_if(0)) \
|
|
in(x:length(x_size) alloc_if(0) free_if(0)) \
|
|
in(ilist:length(0) alloc_if(0) free_if(0)) \
|
|
in(overflow:length(0) alloc_if(0) free_if(0)) \
|
|
in(astart,nthreads,inum,nall,ntypes,vflag,eatom) \
|
|
in(f_stride,nlocal,minlocal,separate_flag,offload) \
|
|
out(f_start:length(f_stride) alloc_if(0) free_if(0)) \
|
|
out(ev_global:length(ev_size) alloc_if(0) free_if(0)) \
|
|
out(timer_compute:length(1) alloc_if(0) free_if(0)) \
|
|
signal(f_start)
|
|
#endif
|
|
{
|
|
#if defined(__MIC__) && defined(_LMP_INTEL_OFFLOAD)
|
|
*timer_compute = MIC_Wtime();
|
|
#endif
|
|
|
|
IP_PRE_repack_for_offload(NEWTON_PAIR, separate_flag, nlocal, nall,
|
|
f_stride, x, 0);
|
|
|
|
acc_t oevdwl, ov0, ov1, ov2, ov3, ov4, ov5;
|
|
if (EFLAG || vflag)
|
|
oevdwl = ov0 = ov1 = ov2 = ov3 = ov4 = ov5 = (acc_t)0;
|
|
if (NEWTON_PAIR == 0 && inum != nlocal)
|
|
memset(f_start, 0, f_stride * sizeof(FORCE_T));
|
|
|
|
// loop over neighbors of my atoms
|
|
#if defined(_OPENMP)
|
|
#pragma omp parallel reduction(+:oevdwl,ov0,ov1,ov2,ov3,ov4,ov5)
|
|
#endif
|
|
{
|
|
int iifrom, iip, iito, tid;
|
|
IP_PRE_omp_stride_id(iifrom, iip, iito, tid, inum, nthreads);
|
|
iifrom += astart;
|
|
iito += astart;
|
|
|
|
int foff;
|
|
if (NEWTON_PAIR) foff = tid * f_stride - minlocal;
|
|
else foff = -minlocal;
|
|
FORCE_T * _noalias const f = f_start + foff;
|
|
if (NEWTON_PAIR) memset(f + minlocal, 0, f_stride * sizeof(FORCE_T));
|
|
|
|
for (int ii = iifrom; ii < iito; ii += iip) {
|
|
const int i = ilist[ii];
|
|
const int itype = x[i].w;
|
|
|
|
const int ptr_off = itype * ntypes;
|
|
const C_FORCE_T * _noalias const c_forcei = c_force + ptr_off;
|
|
const C_ENERGY_T * _noalias const c_energyi = c_energy + ptr_off;
|
|
const int * _noalias const jlist = firstneigh[i];
|
|
int jnum = numneigh[i];
|
|
IP_PRE_neighbor_pad(jnum, offload);
|
|
|
|
acc_t fxtmp,fytmp,fztmp,fwtmp;
|
|
acc_t sevdwl, sv0, sv1, sv2, sv3, sv4, sv5;
|
|
|
|
const flt_t xtmp = x[i].x;
|
|
const flt_t ytmp = x[i].y;
|
|
const flt_t ztmp = x[i].z;
|
|
fxtmp = fytmp = fztmp = (acc_t)0;
|
|
if (EFLAG) fwtmp = sevdwl = (acc_t)0;
|
|
if (NEWTON_PAIR == 0)
|
|
if (vflag == VIRIAL_PAIR)
|
|
sv0 = sv1 = sv2 = sv3 = sv4 = sv5 = (acc_t)0;
|
|
|
|
#if defined(LMP_SIMD_COMPILER)
|
|
#if defined(USE_OMP_SIMD)
|
|
#pragma omp simd reduction(+:fxtmp, fytmp, fztmp, fwtmp, sevdwl, \
|
|
sv0, sv1, sv2, sv3, sv4, sv5)
|
|
#else
|
|
#pragma simd reduction(+:fxtmp, fytmp, fztmp, fwtmp, sevdwl, \
|
|
sv0, sv1, sv2, sv3, sv4, sv5)
|
|
#endif
|
|
#pragma vector aligned
|
|
#endif
|
|
for (int jj = 0; jj < jnum; jj++) {
|
|
|
|
flt_t forcebuck, evdwl;
|
|
forcebuck = evdwl = (flt_t)0.0;
|
|
|
|
const int sbindex = jlist[jj] >> SBBITS & 3;
|
|
const int j = jlist[jj] & NEIGHMASK;
|
|
|
|
const flt_t delx = xtmp - x[j].x;
|
|
const flt_t dely = ytmp - x[j].y;
|
|
const flt_t delz = ztmp - x[j].z;
|
|
const int jtype = IP_PRE_dword_index(x[j].w);
|
|
const flt_t rsq = delx * delx + dely * dely + delz * delz;
|
|
const flt_t r = sqrt(rsq);
|
|
const flt_t r2inv = (flt_t)1.0 / rsq;
|
|
|
|
#ifdef INTEL_VMASK
|
|
if (rsq < c_forcei[jtype].cutsq) {
|
|
#endif
|
|
const flt_t r6inv = r2inv * r2inv * r2inv;
|
|
const flt_t rexp = exp(-r * c_forcei[jtype].rhoinv);
|
|
forcebuck = r * rexp * c_forcei[jtype].buck1 -
|
|
r6inv * c_forcei[jtype].buck2;
|
|
|
|
#ifndef INTEL_VMASK
|
|
if (rsq > c_forcei[jtype].cutsq)
|
|
forcebuck =(flt_t)0.0;
|
|
#endif
|
|
if (EFLAG) {
|
|
evdwl = rexp * c_energyi[jtype].a -
|
|
r6inv * c_energyi[jtype].c -
|
|
c_energyi[jtype].offset;
|
|
|
|
#ifndef INTEL_VMASK
|
|
if (rsq > c_forcei[jtype].cutsq)
|
|
evdwl =(flt_t)0.0;
|
|
#endif
|
|
}
|
|
|
|
if (sbindex) {
|
|
const flt_t factor_lj = special_lj[sbindex];
|
|
forcebuck *= factor_lj;
|
|
if (EFLAG)
|
|
evdwl *= factor_lj;
|
|
}
|
|
const flt_t fpair = forcebuck * r2inv;
|
|
const flt_t fpx = fpair * delx;
|
|
fxtmp += fpx;
|
|
if (NEWTON_PAIR) f[j].x -= fpx;
|
|
const flt_t fpy = fpair * dely;
|
|
fytmp += fpy;
|
|
if (NEWTON_PAIR) f[j].y -= fpy;
|
|
const flt_t fpz = fpair * delz;
|
|
fztmp += fpz;
|
|
if (NEWTON_PAIR) f[j].z -= fpz;
|
|
|
|
if (EFLAG) {
|
|
sevdwl += evdwl;
|
|
if (eatom) {
|
|
fwtmp += (flt_t)0.5 * evdwl;
|
|
if (NEWTON_PAIR)
|
|
f[j].w += (flt_t)0.5 * evdwl;
|
|
}
|
|
}
|
|
if (NEWTON_PAIR == 0)
|
|
IP_PRE_ev_tally_nborv(vflag, delx, dely, delz, fpx, fpy, fpz);
|
|
#ifdef INTEL_VMASK
|
|
}
|
|
#endif
|
|
} // for jj
|
|
if (NEWTON_PAIR) {
|
|
f[i].x += fxtmp;
|
|
f[i].y += fytmp;
|
|
f[i].z += fztmp;
|
|
} else {
|
|
f[i].x = fxtmp;
|
|
f[i].y = fytmp;
|
|
f[i].z = fztmp;
|
|
}
|
|
IP_PRE_ev_tally_atom(NEWTON_PAIR, EFLAG, vflag, f, fwtmp);
|
|
} // for ii
|
|
|
|
IP_PRE_fdotr_reduce_omp(NEWTON_PAIR, nall, minlocal, nthreads, f_start,
|
|
f_stride, x, offload, vflag, ov0, ov1, ov2, ov3,
|
|
ov4, ov5);
|
|
} // end of omp parallel region
|
|
|
|
IP_PRE_fdotr_reduce(NEWTON_PAIR, nall, nthreads, f_stride, vflag,
|
|
ov0, ov1, ov2, ov3, ov4, ov5);
|
|
|
|
if (EFLAG || vflag) {
|
|
if (NEWTON_PAIR == 0) {
|
|
oevdwl *= (acc_t)0.5;
|
|
ov0 *= (acc_t)0.5;
|
|
ov1 *= (acc_t)0.5;
|
|
ov2 *= (acc_t)0.5;
|
|
ov3 *= (acc_t)0.5;
|
|
ov4 *= (acc_t)0.5;
|
|
ov5 *= (acc_t)0.5;
|
|
}
|
|
ev_global[0] = oevdwl;
|
|
ev_global[1] = (acc_t)0;
|
|
ev_global[2] = ov0;
|
|
ev_global[3] = ov1;
|
|
ev_global[4] = ov2;
|
|
ev_global[5] = ov3;
|
|
ev_global[6] = ov4;
|
|
ev_global[7] = ov5;
|
|
}
|
|
#if defined(__MIC__) && defined(_LMP_INTEL_OFFLOAD)
|
|
*timer_compute = MIC_Wtime() - *timer_compute;
|
|
#endif
|
|
} // end of offload region
|
|
|
|
if (offload)
|
|
fix->stop_watch(TIME_OFFLOAD_LATENCY);
|
|
else
|
|
fix->stop_watch(TIME_HOST_PAIR);
|
|
|
|
if (EFLAG || vflag)
|
|
fix->add_result_array(f_start, ev_global, offload, eatom, 0, vflag);
|
|
else
|
|
fix->add_result_array(f_start, nullptr, offload);
|
|
}
|
|
|
|
void PairBuckIntel::init_style()
|
|
{
|
|
PairBuck::init_style();
|
|
|
|
// augment neighbor list request
|
|
if (force->newton_pair == 0)
|
|
neighbor->find_request(this)->enable_full();
|
|
|
|
fix = static_cast<FixIntel *>(modify->get_fix_by_id("package_intel"));
|
|
if (!fix) error->all(FLERR, "The 'package intel' command is required for /intel styles");
|
|
|
|
fix->pair_init_check();
|
|
#ifdef _LMP_INTEL_OFFLOAD
|
|
_cop = fix->coprocessor_number();
|
|
#endif
|
|
|
|
if (fix->precision() == FixIntel::PREC_MODE_MIXED)
|
|
pack_force_const(force_const_single, fix->get_mixed_buffers());
|
|
else if (fix->precision() == FixIntel::PREC_MODE_DOUBLE)
|
|
pack_force_const(force_const_double, fix->get_double_buffers());
|
|
else
|
|
pack_force_const(force_const_single, fix->get_single_buffers());
|
|
}
|
|
|
|
template <class flt_t, class acc_t>
|
|
void PairBuckIntel::pack_force_const(ForceConst<flt_t> &fc,
|
|
IntelBuffers<flt_t,acc_t> *buffers)
|
|
{
|
|
int tp1 = atom->ntypes + 1;
|
|
|
|
fc.set_ntypes(tp1, memory, _cop);
|
|
|
|
// Repeat cutsq calculation because done after call to init_style
|
|
for (int i = 1; i <= atom->ntypes; i++) {
|
|
for (int j = i; j <= atom->ntypes; j++) {
|
|
double cut;
|
|
if (setflag[i][j] != 0 || (setflag[i][i] != 0 && setflag[j][j] != 0))
|
|
cut = init_one(i, j);
|
|
else
|
|
cut = 0.0;
|
|
cutsq[i][j] = cutsq[j][i] = cut*cut;
|
|
}
|
|
}
|
|
|
|
for (int i = 0; i < 4; i++) {
|
|
fc.special_lj[i] = force->special_lj[i];
|
|
fc.special_lj[0] = 1.0;
|
|
}
|
|
|
|
for (int i = 1; i < tp1; i++) {
|
|
for (int j = 1; j < tp1; j++) {
|
|
fc.c_force[i][j].buck1 = buck1[i][j];
|
|
fc.c_force[i][j].buck2 = buck2[i][j];
|
|
fc.c_force[i][j].rhoinv = rhoinv[i][j];
|
|
fc.c_force[i][j].cutsq = cutsq[i][j];
|
|
fc.c_energy[i][j].a = a[i][j];
|
|
fc.c_energy[i][j].c = c[i][j];
|
|
fc.c_energy[i][j].offset = offset[i][j];
|
|
}
|
|
}
|
|
|
|
#ifdef _LMP_INTEL_OFFLOAD
|
|
if (_cop < 0) return;
|
|
flt_t * special_lj = fc.special_lj;
|
|
C_FORCE_T * c_force = fc.c_force[0];
|
|
C_ENERGY_T * c_energy = fc.c_energy[0];
|
|
int tp1sq = tp1 * tp1;
|
|
#pragma offload_transfer target(mic:_cop) \
|
|
in(special_lj: length(4) alloc_if(0) free_if(0)) \
|
|
in(c_force, c_energy: length(tp1sq) alloc_if(0) free_if(0))
|
|
#endif
|
|
}
|
|
|
|
/* ---------------------------------------------------------------------- */
|
|
|
|
template <class flt_t>
|
|
void PairBuckIntel::ForceConst<flt_t>::set_ntypes(const int ntypes,
|
|
Memory *memory,
|
|
const int cop) {
|
|
if (memory != nullptr) _memory = memory;
|
|
if ((ntypes != _ntypes )) {
|
|
if (_ntypes > 0) {
|
|
#ifdef _LMP_INTEL_OFFLOAD
|
|
flt_t * ospecial_lj = special_lj;
|
|
c_force_t * oc_force = c_force[0];
|
|
c_energy_t * oc_energy = c_energy[0];
|
|
|
|
if (ospecial_lj != nullptr && oc_force != nullptr &&
|
|
oc_energy != nullptr &&
|
|
_cop >= 0) {
|
|
#pragma offload_transfer target(mic:cop) \
|
|
nocopy(ospecial_lj: alloc_if(0) free_if(1)) \
|
|
nocopy(oc_force, oc_energy: alloc_if(0) free_if(1))
|
|
}
|
|
#endif
|
|
|
|
_memory->destroy(c_force);
|
|
_memory->destroy(c_energy);
|
|
|
|
}
|
|
if (ntypes > 0) {
|
|
_cop = cop;
|
|
_memory->create(c_force,ntypes,ntypes,"fc.c_force");
|
|
_memory->create(c_energy,ntypes,ntypes,"fc.c_energy");
|
|
|
|
#ifdef _LMP_INTEL_OFFLOAD
|
|
flt_t * ospecial_lj = special_lj;
|
|
c_force_t * oc_force = c_force[0];
|
|
c_energy_t * oc_energy = c_energy[0];
|
|
int tp1sq = ntypes*ntypes;
|
|
if (ospecial_lj != nullptr && oc_force != nullptr &&
|
|
oc_energy != nullptr &&
|
|
cop >= 0) {
|
|
#pragma offload_transfer target(mic:cop) \
|
|
nocopy(ospecial_lj: length(4) alloc_if(1) free_if(0)) \
|
|
nocopy(oc_force: length(tp1sq) alloc_if(1) free_if(0)) \
|
|
nocopy(oc_energy: length(tp1sq) alloc_if(1) free_if(0))
|
|
|
|
}
|
|
#endif
|
|
}
|
|
}
|
|
_ntypes=ntypes;
|
|
}
|
|
|
|
|