408 lines
18 KiB
C++
408 lines
18 KiB
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: Trung Dac Nguyen (ORNL)
|
||
------------------------------------------------------------------------- */
|
||
|
||
#include "pair_dpd_tstat_gpu.h"
|
||
|
||
#include "atom.h"
|
||
#include "domain.h"
|
||
#include "error.h"
|
||
#include "force.h"
|
||
#include "gpu_extra.h"
|
||
#include "neigh_list.h"
|
||
#include "neighbor.h"
|
||
#include "suffix.h"
|
||
#include "update.h"
|
||
|
||
#include <cmath>
|
||
|
||
using namespace LAMMPS_NS;
|
||
|
||
// External functions from cuda library for atom decomposition
|
||
|
||
int dpd_tstat_gpu_init(const int ntypes, double **cutsq, double **host_a0, double **host_gamma,
|
||
double **host_sigma, double **host_cut, double *special_lj, const int inum,
|
||
const int nall, const int max_nbors, const int maxspecial,
|
||
const double cell_size, int &gpu_mode, FILE *screen);
|
||
void dpd_tstat_gpu_clear();
|
||
int **dpd_tstat_gpu_compute_n(const int ago, const int inum_full, const int nall, double **host_x,
|
||
int *host_type, double *sublo, double *subhi, tagint *tag,
|
||
int **nspecial, tagint **special, const bool eflag, const bool vflag,
|
||
const bool eatom, const bool vatom, int &host_start, int **ilist,
|
||
int **jnum, const double cpu_time, bool &success, double **host_v,
|
||
const double dtinvsqrt, const int seed, const int timestep,
|
||
double *boxlo, double *prd);
|
||
void dpd_tstat_gpu_compute(const int ago, const int inum_full, const int nall, double **host_x,
|
||
int *host_type, int *ilist, int *numj, int **firstneigh,
|
||
const bool eflag, const bool vflag, const bool eatom, const bool vatom,
|
||
int &host_start, const double cpu_time, bool &success, tagint *tag,
|
||
double **host_v, const double dtinvsqrt, const int seed,
|
||
const int timestep, const int nlocal, double *boxlo, double *prd);
|
||
void dpd_tstat_gpu_update_coeff(int ntypes, double **host_a0, double **host_gamma,
|
||
double **host_sigma, double **host_cut);
|
||
double dpd_tstat_gpu_bytes();
|
||
|
||
#define EPSILON 1.0e-10
|
||
|
||
//#define _USE_UNIFORM_SARU_LCG
|
||
//#define _USE_UNIFORM_SARU_TEA8
|
||
//#define _USE_GAUSSIAN_SARU_LCG
|
||
|
||
#if !defined(_USE_UNIFORM_SARU_LCG) && !defined(_USE_UNIFORM_SARU_TEA8) && \
|
||
!defined(_USE_GAUSSIAN_SARU_LCG)
|
||
#define _USE_UNIFORM_SARU_LCG
|
||
#endif
|
||
|
||
// References:
|
||
// 1. Y. Afshar, F. Schmid, A. Pishevar, S. Worley, Comput. Phys. Comm. 184 (2013), 1119–1128.
|
||
// 2. C. L. Phillips, J. A. Anderson, S. C. Glotzer, Comput. Phys. Comm. 230 (2011), 7191-7201.
|
||
// PRNG period = 3666320093*2^32 ~ 2^64 ~ 10^19
|
||
|
||
#define LCGA 0x4beb5d59 // Full period 32 bit LCG
|
||
#define LCGC 0x2600e1f7
|
||
#define oWeylPeriod 0xda879add // Prime period 3666320093
|
||
#define oWeylOffset 0x8009d14b
|
||
#define TWO_N32 0.232830643653869628906250e-9f /* 2^-32 */
|
||
|
||
// specifically implemented for steps = 1; high = 1.0; low = -1.0
|
||
// returns uniformly distributed random numbers u in [-1.0;1.0]
|
||
// using the inherent LCG, then multiply u with sqrt(3) to "match"
|
||
// with a normal random distribution.
|
||
// Afshar et al. mutlplies u in [-0.5;0.5] with sqrt(12)
|
||
// Curly brackets to make variables local to the scope.
|
||
#ifdef _USE_UNIFORM_SARU_LCG
|
||
#define numtyp double
|
||
#define SQRT3 (numtyp) 1.7320508075688772935274463
|
||
#define saru(seed1, seed2, seed, timestep, randnum) \
|
||
{ \
|
||
unsigned int seed3 = seed + timestep; \
|
||
seed3 ^= (seed1 << 7) ^ (seed2 >> 6); \
|
||
seed2 += (seed1 >> 4) ^ (seed3 >> 15); \
|
||
seed1 ^= (seed2 << 9) + (seed3 << 8); \
|
||
seed3 ^= 0xA5366B4D * ((seed2 >> 11) ^ (seed1 << 1)); \
|
||
seed2 += 0x72BE1579 * ((seed1 << 4) ^ (seed3 >> 16)); \
|
||
seed1 ^= 0x3F38A6ED * ((seed3 >> 5) ^ (((signed int) seed2) >> 22)); \
|
||
seed2 += seed1 * seed3; \
|
||
seed1 += seed3 ^ (seed2 >> 2); \
|
||
seed2 ^= ((signed int) seed2) >> 17; \
|
||
unsigned int state = 0x79dedea3 * (seed1 ^ (((signed int) seed1) >> 14)); \
|
||
unsigned int wstate = (state + seed2) ^ (((signed int) state) >> 8); \
|
||
state = state + (wstate * (wstate ^ 0xdddf97f5)); \
|
||
wstate = 0xABCB96F7 + (wstate >> 1); \
|
||
state = LCGA * state + LCGC; \
|
||
wstate = wstate + oWeylOffset + ((((signed int) wstate) >> 31) & oWeylPeriod); \
|
||
unsigned int v = (state ^ (state >> 26)) + wstate; \
|
||
unsigned int s = (signed int) ((v ^ (v >> 20)) * 0x6957f5a7); \
|
||
randnum = SQRT3 * (s * TWO_N32 * (numtyp) 2.0 - (numtyp) 1.0); \
|
||
}
|
||
#endif
|
||
|
||
// specifically implemented for steps = 1; high = 1.0; low = -1.0
|
||
// returns uniformly distributed random numbers u in [-1.0;1.0] using TEA8
|
||
// then multiply u with sqrt(3) to "match" with a normal random distribution
|
||
// Afshar et al. mutlplies u in [-0.5;0.5] with sqrt(12)
|
||
#ifdef _USE_UNIFORM_SARU_TEA8
|
||
#define numtyp double
|
||
#define SQRT3 (numtyp) 1.7320508075688772935274463
|
||
#define k0 0xA341316C
|
||
#define k1 0xC8013EA4
|
||
#define k2 0xAD90777D
|
||
#define k3 0x7E95761E
|
||
#define delta 0x9e3779b9
|
||
#define rounds 8
|
||
#define saru(seed1, seed2, seed, timestep, randnum) \
|
||
{ \
|
||
unsigned int seed3 = seed + timestep; \
|
||
seed3 ^= (seed1 << 7) ^ (seed2 >> 6); \
|
||
seed2 += (seed1 >> 4) ^ (seed3 >> 15); \
|
||
seed1 ^= (seed2 << 9) + (seed3 << 8); \
|
||
seed3 ^= 0xA5366B4D * ((seed2 >> 11) ^ (seed1 << 1)); \
|
||
seed2 += 0x72BE1579 * ((seed1 << 4) ^ (seed3 >> 16)); \
|
||
seed1 ^= 0x3F38A6ED * ((seed3 >> 5) ^ (((signed int) seed2) >> 22)); \
|
||
seed2 += seed1 * seed3; \
|
||
seed1 += seed3 ^ (seed2 >> 2); \
|
||
seed2 ^= ((signed int) seed2) >> 17; \
|
||
unsigned int state = 0x79dedea3 * (seed1 ^ (((signed int) seed1) >> 14)); \
|
||
unsigned int wstate = (state + seed2) ^ (((signed int) state) >> 8); \
|
||
state = state + (wstate * (wstate ^ 0xdddf97f5)); \
|
||
wstate = 0xABCB96F7 + (wstate >> 1); \
|
||
unsigned int sum = 0; \
|
||
for (int i = 0; i < rounds; i++) { \
|
||
sum += delta; \
|
||
state += ((wstate << 4) + k0) ^ (wstate + sum) ^ ((wstate >> 5) + k1); \
|
||
wstate += ((state << 4) + k2) ^ (state + sum) ^ ((state >> 5) + k3); \
|
||
} \
|
||
unsigned int v = (state ^ (state >> 26)) + wstate; \
|
||
unsigned int s = (signed int) ((v ^ (v >> 20)) * 0x6957f5a7); \
|
||
randnum = SQRT3 * (s * TWO_N32 * (numtyp) 2.0 - (numtyp) 1.0); \
|
||
}
|
||
#endif
|
||
|
||
// specifically implemented for steps = 1; high = 1.0; low = -1.0
|
||
// returns two uniformly distributed random numbers r1 and r2 in [-1.0;1.0],
|
||
// and uses the polar method (Marsaglia's) to transform to a normal random value
|
||
// This is used to compared with CPU DPD using RandMars::gaussian()
|
||
#ifdef _USE_GAUSSIAN_SARU_LCG
|
||
#define numtyp double
|
||
#define saru(seed1, seed2, seed, timestep, randnum) \
|
||
{ \
|
||
unsigned int seed3 = seed + timestep; \
|
||
seed3 ^= (seed1 << 7) ^ (seed2 >> 6); \
|
||
seed2 += (seed1 >> 4) ^ (seed3 >> 15); \
|
||
seed1 ^= (seed2 << 9) + (seed3 << 8); \
|
||
seed3 ^= 0xA5366B4D * ((seed2 >> 11) ^ (seed1 << 1)); \
|
||
seed2 += 0x72BE1579 * ((seed1 << 4) ^ (seed3 >> 16)); \
|
||
seed1 ^= 0x3F38A6ED * ((seed3 >> 5) ^ (((signed int) seed2) >> 22)); \
|
||
seed2 += seed1 * seed3; \
|
||
seed1 += seed3 ^ (seed2 >> 2); \
|
||
seed2 ^= ((signed int) seed2) >> 17; \
|
||
unsigned int state = 0x12345678; \
|
||
unsigned int wstate = 12345678; \
|
||
state = 0x79dedea3 * (seed1 ^ (((signed int) seed1) >> 14)); \
|
||
wstate = (state + seed2) ^ (((signed int) state) >> 8); \
|
||
state = state + (wstate * (wstate ^ 0xdddf97f5)); \
|
||
wstate = 0xABCB96F7 + (wstate >> 1); \
|
||
unsigned int v, s; \
|
||
numtyp r1, r2, rsq; \
|
||
while (1) { \
|
||
state = LCGA * state + LCGC; \
|
||
wstate = wstate + oWeylOffset + ((((signed int) wstate) >> 31) & oWeylPeriod); \
|
||
v = (state ^ (state >> 26)) + wstate; \
|
||
s = (signed int) ((v ^ (v >> 20)) * 0x6957f5a7); \
|
||
r1 = s * TWO_N32 * (numtyp) 2.0 - (numtyp) 1.0; \
|
||
state = LCGA * state + LCGC; \
|
||
wstate = wstate + oWeylOffset + ((((signed int) wstate) >> 31) & oWeylPeriod); \
|
||
v = (state ^ (state >> 26)) + wstate; \
|
||
s = (signed int) ((v ^ (v >> 20)) * 0x6957f5a7); \
|
||
r2 = s * TWO_N32 * (numtyp) 2.0 - (numtyp) 1.0; \
|
||
rsq = r1 * r1 + r2 * r2; \
|
||
if (rsq < (numtyp) 1.0) break; \
|
||
} \
|
||
numtyp fac = sqrt((numtyp) -2.0 * log(rsq) / rsq); \
|
||
randnum = r2 * fac; \
|
||
}
|
||
#endif
|
||
|
||
/* ---------------------------------------------------------------------- */
|
||
|
||
PairDPDTstatGPU::PairDPDTstatGPU(LAMMPS *lmp) : PairDPDTstat(lmp), gpu_mode(GPU_FORCE)
|
||
{
|
||
respa_enable = 0;
|
||
reinitflag = 0;
|
||
cpu_time = 0.0;
|
||
suffix_flag |= Suffix::GPU;
|
||
GPU_EXTRA::gpu_ready(lmp->modify, lmp->error);
|
||
}
|
||
|
||
/* ----------------------------------------------------------------------
|
||
free all arrays
|
||
------------------------------------------------------------------------- */
|
||
|
||
PairDPDTstatGPU::~PairDPDTstatGPU()
|
||
{
|
||
dpd_tstat_gpu_clear();
|
||
}
|
||
|
||
/* ---------------------------------------------------------------------- */
|
||
|
||
void PairDPDTstatGPU::compute(int eflag, int vflag)
|
||
{
|
||
ev_init(eflag, vflag);
|
||
|
||
// adjust sigma if target T is changing
|
||
|
||
if (t_start != t_stop) {
|
||
double delta = update->ntimestep - update->beginstep;
|
||
if (delta != 0.0) delta /= update->endstep - update->beginstep;
|
||
temperature = t_start + delta * (t_stop - t_start);
|
||
double boltz = force->boltz;
|
||
for (int i = 1; i <= atom->ntypes; i++)
|
||
for (int j = i; j <= atom->ntypes; j++)
|
||
sigma[i][j] = sigma[j][i] = sqrt(2.0 * boltz * temperature * gamma[i][j]);
|
||
|
||
dpd_tstat_gpu_update_coeff(atom->ntypes + 1, a0, gamma, sigma, cut);
|
||
}
|
||
|
||
int nall = atom->nlocal + atom->nghost;
|
||
int inum, host_start;
|
||
|
||
double dtinvsqrt = 1.0 / sqrt(update->dt);
|
||
|
||
bool success = true;
|
||
int *ilist, *numneigh, **firstneigh;
|
||
if (gpu_mode != GPU_FORCE) {
|
||
double sublo[3], subhi[3];
|
||
if (domain->triclinic == 0) {
|
||
sublo[0] = domain->sublo[0];
|
||
sublo[1] = domain->sublo[1];
|
||
sublo[2] = domain->sublo[2];
|
||
subhi[0] = domain->subhi[0];
|
||
subhi[1] = domain->subhi[1];
|
||
subhi[2] = domain->subhi[2];
|
||
} else {
|
||
domain->bbox(domain->sublo_lamda, domain->subhi_lamda, sublo, subhi);
|
||
}
|
||
inum = atom->nlocal;
|
||
firstneigh = dpd_tstat_gpu_compute_n(
|
||
neighbor->ago, inum, nall, atom->x, atom->type, sublo, subhi, atom->tag, atom->nspecial,
|
||
atom->special, eflag, vflag, eflag_atom, vflag_atom, host_start, &ilist, &numneigh,
|
||
cpu_time, success, atom->v, dtinvsqrt, seed, update->ntimestep, domain->boxlo, domain->prd);
|
||
} else {
|
||
inum = list->inum;
|
||
ilist = list->ilist;
|
||
numneigh = list->numneigh;
|
||
firstneigh = list->firstneigh;
|
||
dpd_tstat_gpu_compute(neighbor->ago, inum, nall, atom->x, atom->type, ilist, numneigh,
|
||
firstneigh, eflag, vflag, eflag_atom, vflag_atom, host_start, cpu_time,
|
||
success, atom->tag, atom->v, dtinvsqrt, seed, update->ntimestep,
|
||
atom->nlocal, domain->boxlo, domain->prd);
|
||
}
|
||
if (!success) error->one(FLERR, "Insufficient memory on accelerator");
|
||
|
||
if (host_start < inum) {
|
||
cpu_time = platform::walltime();
|
||
cpu_compute(host_start, inum, eflag, vflag, ilist, numneigh, firstneigh);
|
||
cpu_time = platform::walltime() - cpu_time;
|
||
}
|
||
}
|
||
|
||
/* ----------------------------------------------------------------------
|
||
init specific to this pair style
|
||
------------------------------------------------------------------------- */
|
||
|
||
void PairDPDTstatGPU::init_style()
|
||
{
|
||
|
||
// Repeat cutsq calculation because done after call to init_style
|
||
double maxcut = -1.0;
|
||
double mcut;
|
||
for (int i = 1; i <= atom->ntypes; i++) {
|
||
for (int j = i; j <= atom->ntypes; j++) {
|
||
if (setflag[i][j] != 0 || (setflag[i][i] != 0 && setflag[j][j] != 0)) {
|
||
mcut = init_one(i, j);
|
||
mcut *= mcut;
|
||
if (mcut > maxcut) maxcut = mcut;
|
||
cutsq[i][j] = cutsq[j][i] = mcut;
|
||
} else
|
||
cutsq[i][j] = cutsq[j][i] = 0.0;
|
||
}
|
||
}
|
||
double cell_size = sqrt(maxcut) + neighbor->skin;
|
||
|
||
int maxspecial = 0;
|
||
if (atom->molecular != Atom::ATOMIC) maxspecial = atom->maxspecial;
|
||
int mnf = 5e-2 * neighbor->oneatom;
|
||
int success = dpd_tstat_gpu_init(atom->ntypes + 1, cutsq, a0, gamma, sigma, cut,
|
||
force->special_lj, atom->nlocal, atom->nlocal + atom->nghost,
|
||
mnf, maxspecial, cell_size, gpu_mode, screen);
|
||
GPU_EXTRA::check_flag(success, error, world);
|
||
|
||
if (gpu_mode == GPU_FORCE) neighbor->add_request(this, NeighConst::REQ_FULL);
|
||
}
|
||
|
||
/* ---------------------------------------------------------------------- */
|
||
|
||
double PairDPDTstatGPU::memory_usage()
|
||
{
|
||
double bytes = Pair::memory_usage();
|
||
return bytes + dpd_tstat_gpu_bytes();
|
||
}
|
||
|
||
/* ---------------------------------------------------------------------- */
|
||
|
||
void PairDPDTstatGPU::cpu_compute(int start, int inum, int /* eflag */, int /* vflag */, int *ilist,
|
||
int *numneigh, int **firstneigh)
|
||
{
|
||
int i, j, ii, jj, jnum, itype, jtype;
|
||
double xtmp, ytmp, ztmp, delx, dely, delz, fpair;
|
||
double vxtmp, vytmp, vztmp, delvx, delvy, delvz;
|
||
double rsq, r, rinv, dot, wd, randnum, factor_dpd;
|
||
int *jlist;
|
||
tagint itag, jtag;
|
||
|
||
double **x = atom->x;
|
||
double **v = atom->v;
|
||
double **f = atom->f;
|
||
int *type = atom->type;
|
||
tagint *tag = atom->tag;
|
||
double *special_lj = force->special_lj;
|
||
double dtinvsqrt = 1.0 / sqrt(update->dt);
|
||
int timestep = (int) update->ntimestep;
|
||
|
||
// loop over neighbors of my atoms
|
||
|
||
for (ii = start; ii < inum; ii++) {
|
||
i = ilist[ii];
|
||
xtmp = x[i][0];
|
||
ytmp = x[i][1];
|
||
ztmp = x[i][2];
|
||
vxtmp = v[i][0];
|
||
vytmp = v[i][1];
|
||
vztmp = v[i][2];
|
||
itype = type[i];
|
||
itag = tag[i];
|
||
jlist = firstneigh[i];
|
||
jnum = numneigh[i];
|
||
|
||
for (jj = 0; jj < jnum; jj++) {
|
||
j = jlist[jj];
|
||
factor_dpd = special_lj[sbmask(j)];
|
||
j &= NEIGHMASK;
|
||
|
||
delx = xtmp - x[j][0];
|
||
dely = ytmp - x[j][1];
|
||
delz = ztmp - x[j][2];
|
||
rsq = delx * delx + dely * dely + delz * delz;
|
||
jtype = type[j];
|
||
jtag = tag[j];
|
||
|
||
if (rsq < cutsq[itype][jtype]) {
|
||
r = sqrt(rsq);
|
||
if (r < EPSILON) continue; // r can be 0.0 in DPD systems
|
||
rinv = 1.0 / r;
|
||
delvx = vxtmp - v[j][0];
|
||
delvy = vytmp - v[j][1];
|
||
delvz = vztmp - v[j][2];
|
||
dot = delx * delvx + dely * delvy + delz * delvz;
|
||
wd = 1.0 - r / cut[itype][jtype];
|
||
|
||
unsigned int tag1 = itag, tag2 = jtag;
|
||
if (tag1 > tag2) {
|
||
tag1 = jtag;
|
||
tag2 = itag;
|
||
}
|
||
|
||
randnum = 0.0;
|
||
saru(tag1, tag2, seed, timestep, randnum);
|
||
|
||
// conservative force = a0 * wd
|
||
// drag force = -gamma * wd^2 * (delx dot delv) / r
|
||
// random force = sigma * wd * rnd * dtinvsqrt;
|
||
|
||
fpair = -gamma[itype][jtype] * wd * wd * dot * rinv;
|
||
fpair += sigma[itype][jtype] * wd * randnum * dtinvsqrt;
|
||
fpair *= factor_dpd * rinv;
|
||
|
||
f[i][0] += delx * fpair;
|
||
f[i][1] += dely * fpair;
|
||
f[i][2] += delz * fpair;
|
||
|
||
if (evflag) ev_tally_full(i, 0.0, 0.0, fpair, delx, dely, delz);
|
||
}
|
||
}
|
||
}
|
||
}
|