ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
b093f1aac19f056e0decfccef23f95d24cd6d44f
lammps/lib/gpu/lal_coul_long_cs.cu
W. Michael Brown 37f22c8627 Misc Improvements to GPU Package 
- Optimizations for molecular systems
-   Improved kernel performance and greater CPU overlap
- Reduced GPU to CPU communications for discrete devices
- Switch classic Intel makefiles to use LLVM-based compilers
- Prefetch optimizations supported for OpenCL
- Optimized data repack for quaternions
2023-03-05 21:03:12 -08:00

285 lines
9.6 KiB
Plaintext
Raw Blame History

// **************************************************************************
// coul_long_cs.cu
// -------------------
// Trung Nguyen (Northwestern)
//
// Device code for acceleration of the coul/long/cs pair style
//
// __________________________________________________________________________
// This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
// __________________________________________________________________________
//
// begin : June 2018
// email : ndactrung@gmail.com
// ***************************************************************************
#if defined(NV_KERNEL) || defined(USE_HIP)
#include "lal_aux_fun1.h"
#ifndef _DOUBLE_DOUBLE
_texture( pos_tex,float4);
_texture( q_tex,float);
#else
_texture_2d( pos_tex,int4);
_texture( q_tex,int2);
#endif
#else
#define pos_tex x_
#define q_tex q_
#endif
// Note: EWALD_P is different from that in lal_preprocessor.h
// acctyp is needed for these parameters
#define CS_EWALD_P (acctyp)9.95473818e-1
#define B0 (acctyp)-0.1335096380159268
#define B1 (acctyp)-2.57839507e-1
#define B2 (acctyp)-1.37203639e-1
#define B3 (acctyp)-8.88822059e-3
#define B4 (acctyp)-5.80844129e-3
#define B5 (acctyp)1.14652755e-1
#define EPSILON (acctyp)(1.0e-20)
#define EPS_EWALD (acctyp)(1.0e-6)
#define EPS_EWALD_SQR (acctyp)(1.0e-12)
__kernel void k_coul_long_cs(const __global numtyp4 *restrict x_,
const __global numtyp *restrict scale,
const int lj_types,
const __global numtyp *restrict sp_cl_in,
const __global int *dev_nbor,
const __global int *dev_packed,
__global acctyp3 *restrict ans,
__global acctyp *restrict engv,
const int eflag, const int vflag, const int inum,
const int nbor_pitch,
const __global numtyp *restrict q_,
const numtyp cut_coulsq, const numtyp qqrd2e,
const numtyp g_ewald, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_cl[4];
int n_stride;
local_allocate_store_charge();
sp_cl[0]=sp_cl_in[0];
sp_cl[1]=sp_cl_in[1];
sp_cl[2]=sp_cl_in[2];
sp_cl[3]=sp_cl_in[3];
acctyp3 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp e_coul, virial[6];
if (EVFLAG) {
e_coul=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
if (ii<inum) {
int nbor, nbor_end;
int i, numj;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
int itype=ix.w;
numtyp qtmp; fetch(qtmp,i,q_tex);
for ( ; nbor<nbor_end; nbor+=n_stride) {
ucl_prefetch(dev_packed+nbor+n_stride);
int j=dev_packed[nbor];
numtyp factor_coul;
factor_coul = sp_cl[sbmask(j)];
j &= NEIGHMASK;
numtyp4 jx; fetch4(jx,j,pos_tex); //x_[j];
int jtype=jx.w;
// Compute r12
numtyp delx = ix.x-jx.x;
numtyp dely = ix.y-jx.y;
numtyp delz = ix.z-jx.z;
numtyp rsq = delx*delx+dely*dely+delz*delz;
int mtype=itype*lj_types+jtype;
if (rsq < cut_coulsq) {
rsq += EPSILON; // Add Epsilon for case: r = 0; Interaction must be removed by special bond;
numtyp force,prefactor,_erfc;
numtyp r2inv = ucl_recip(rsq);
numtyp r = ucl_rsqrt(r2inv);
fetch(prefactor,j,q_tex);
prefactor *= qqrd2e * scale[mtype] * qtmp;
if (factor_coul<(numtyp)1.0) {
numtyp grij = g_ewald * (r+EPS_EWALD);
numtyp expm2 = ucl_exp(-grij*grij);
acctyp t = ucl_recip((numtyp)1.0 + CS_EWALD_P*grij);
numtyp u = (numtyp)1.0 - t;
_erfc = t * ((numtyp)1.0 + u*(B0+u*(B1+u*(B2+u*(B3+u*(B4+u*B5)))))) * expm2;
prefactor /= (r+EPS_EWALD);
force = prefactor * (_erfc + EWALD_F*grij*expm2 - ((numtyp)1.0-factor_coul));
// Additionally r2inv needs to be accordingly modified since the later
// scaling of the overall force shall be consistent
r2inv = ucl_recip(rsq + EPS_EWALD_SQR);
force *= r2inv;
} else {
numtyp grij = g_ewald * r;
numtyp expm2 = ucl_exp(-grij*grij);
acctyp t = ucl_recip((numtyp)1.0 + CS_EWALD_P*grij);
numtyp u = (numtyp)1.0 - t;
_erfc = t * ((numtyp)1.0 + u*(B0+u*(B1+u*(B2+u*(B3+u*(B4+u*B5)))))) * expm2;
prefactor /= r;
force = prefactor*(_erfc + EWALD_F*grij*expm2);
force *= r2inv;
}
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (EVFLAG && eflag) {
numtyp e = prefactor*_erfc;
if (factor_coul<(numtyp)1.0) e -= ((numtyp)1.0-factor_coul)*prefactor;
e_coul += e;
}
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
virial[3] += delx*dely*force;
virial[4] += delx*delz*force;
virial[5] += dely*delz*force;
}
}
} // for nbor
} // if ii
acctyp energy;
if (EVFLAG) energy=(acctyp)0.0;
store_answers_q(f,energy,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
}
__kernel void k_coul_long_cs_fast(const __global numtyp4 *restrict x_,
const __global numtyp *restrict scale_in,
const __global numtyp *restrict sp_cl_in,
const __global int *dev_nbor,
const __global int *dev_packed,
__global acctyp3 *restrict ans,
__global acctyp *restrict engv,
const int eflag, const int vflag, const int inum,
const int nbor_pitch,
const __global numtyp *restrict q_,
const numtyp cut_coulsq, const numtyp qqrd2e,
const numtyp g_ewald, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp scale[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_cl[4];
int n_stride;
local_allocate_store_charge();
if (tid<4)
sp_cl[tid]=sp_cl_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES)
scale[tid]=scale_in[tid];
acctyp3 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp e_coul, virial[6];
if (EVFLAG) {
e_coul=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
if (ii<inum) {
int nbor, nbor_end;
int i, numj;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
numtyp qtmp; fetch(qtmp,i,q_tex);
int iw=ix.w;
int itype=fast_mul((int)MAX_SHARED_TYPES,iw);
for ( ; nbor<nbor_end; nbor+=n_stride) {
ucl_prefetch(dev_packed+nbor+n_stride);
int j=dev_packed[nbor];
numtyp factor_coul;
factor_coul = sp_cl[sbmask(j)];
j &= NEIGHMASK;
numtyp4 jx; fetch4(jx,j,pos_tex); //x_[j];
int mtype=itype+jx.w;
// Compute r12
numtyp delx = ix.x-jx.x;
numtyp dely = ix.y-jx.y;
numtyp delz = ix.z-jx.z;
numtyp rsq = delx*delx+dely*dely+delz*delz;
if (rsq < cut_coulsq) {
rsq += EPSILON; // Add Epsilon for case: r = 0; Interaction must be removed by special bond;
numtyp force,prefactor,_erfc;
numtyp r2inv = ucl_recip(rsq);
numtyp r = ucl_rsqrt(r2inv);
fetch(prefactor,j,q_tex);
prefactor *= qqrd2e * scale[mtype] * qtmp;
if (factor_coul<(numtyp)1.0) {
numtyp grij = g_ewald * (r+EPS_EWALD);
numtyp expm2 = ucl_exp(-grij*grij);
acctyp t = ucl_recip((numtyp)1.0 + CS_EWALD_P*grij);
numtyp u = (numtyp)1.0 - t;
_erfc = t * ((numtyp)1.0 + u*(B0+u*(B1+u*(B2+u*(B3+u*(B4+u*B5)))))) * expm2;
prefactor /= (r+EPS_EWALD);
force = prefactor * (_erfc + EWALD_F*grij*expm2 - ((numtyp)1.0-factor_coul));
// Additionally r2inv needs to be accordingly modified since the later
// scaling of the overall force shall be consistent
r2inv = ucl_recip(rsq + EPS_EWALD_SQR);
} else {
numtyp grij = g_ewald * r;
numtyp expm2 = ucl_exp(-grij*grij);
acctyp t = ucl_recip((numtyp)1.0 + CS_EWALD_P*grij);
numtyp u = (numtyp)1.0 - t;
_erfc = t * ((numtyp)1.0 + u*(B0+u*(B1+u*(B2+u*(B3+u*(B4+u*B5)))))) * expm2;
prefactor /= r;
force = prefactor * (_erfc + EWALD_F*grij*expm2);
}
force *= r2inv;
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (EVFLAG && eflag) {
numtyp e = prefactor*_erfc;
if (factor_coul<(numtyp)1.0) e -= ((numtyp)1.0-factor_coul)*prefactor;
e_coul += e;
}
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
virial[3] += delx*dely*force;
virial[4] += delx*delz*force;
virial[5] += dely*delz*force;
}
}
} // for nbor
} // if ii
acctyp energy;
if (EVFLAG) energy=(acctyp)0.0;
store_answers_q(f,energy,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
}
Reference in New Issue View Git Blame Copy Permalink