ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

git-svn-id: svn://svn.icms.temple.edu/lammps-ro/trunk@8694 f3b2605a-c512-4ea7-a41b-209d697bcdaa

Browse Source
This commit is contained in:
sjplimp
2012-08-21 13:59:15 +00:00
parent 31551d81fd
commit 647ea4c29f
49 changed files with 8851 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

189
lib/gpu/lal_gauss.cu Normal file
View File

@ -0,0 +1,189 @@
// **************************************************************************
// gauss.cu
// -------------------
// Trung Dac Nguyen (ORNL)
//
// Device code for acceleration of the gauss pair style
//
// __________________________________________________________________________
// This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
// __________________________________________________________________________
//
// begin :
// email : nguyentd@ornl.gov
// ***************************************************************************/
#ifdef NV_KERNEL
#include "lal_aux_fun1.h"
#ifndef _DOUBLE_DOUBLE
texture<float4> pos_tex;
#else
texture<int4,1> pos_tex;
#endif
#else
#define pos_tex x_
#endif
__kernel void k_gauss(__global numtyp4 *x_, __global numtyp4 *gauss1,
const int lj_types,
__global numtyp *sp_lj_in, __global int *dev_nbor,
__global int *dev_packed, __global acctyp4 *ans,
__global acctyp *engv, const int eflag,
const int vflag, const int inum,
const int nbor_pitch, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_lj[4];
sp_lj[0]=sp_lj_in[0];
sp_lj[1]=sp_lj_in[1];
sp_lj[2]=sp_lj_in[2];
sp_lj[3]=sp_lj_in[3];
acctyp energy=(acctyp)0;
acctyp4 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp virial[6];
for (int i=0; i<6; i++)
virial[i]=(acctyp)0;
if (ii<inum) {
__global int *nbor, *list_end;
int i, numj, n_stride;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,list_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
int itype=ix.w;
numtyp factor_lj;
for ( ; nbor<list_end; nbor+=n_stride) {
int j=*nbor;
factor_lj = sp_lj[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<gauss1[mtype].z) {
numtyp r2inv = ucl_recip(rsq);
numtyp r = ucl_sqrt(rsq);
numtyp force = (numtyp)-2.0*gauss1[mtype].x*gauss1[mtype].y*rsq*
ucl_exp(-gauss1[mtype].y*rsq)*r2inv*factor_lj;
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
numtyp e=-(gauss1[mtype].x*ucl_exp(-gauss1[mtype].y*rsq) -
gauss1[mtype].w);
energy+=factor_lj*e;
}
if (vflag>0) {
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
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
}
__kernel void k_gauss_fast(__global numtyp4 *x_, __global numtyp4 *gauss1_in,
__global numtyp* sp_lj_in,
__global int *dev_nbor, __global int *dev_packed,
__global acctyp4 *ans, __global acctyp *engv,
const int eflag, const int vflag, const int inum,
const int nbor_pitch, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp4 gauss1[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_lj[4];
if (tid<4)
sp_lj[tid]=sp_lj_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
gauss1[tid]=gauss1_in[tid];
}
acctyp energy=(acctyp)0;
acctyp4 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp virial[6];
for (int i=0; i<6; i++)
virial[i]=(acctyp)0;
__syncthreads();
if (ii<inum) {
__global int *nbor, *list_end;
int i, numj, n_stride;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,list_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
int iw=ix.w;
int itype=fast_mul((int)MAX_SHARED_TYPES,iw);
numtyp factor_lj;
for ( ; nbor<list_end; nbor+=n_stride) {
int j=*nbor;
factor_lj = sp_lj[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<gauss1[mtype].z) {
numtyp r2inv = ucl_recip(rsq);
numtyp r = ucl_sqrt(rsq);
numtyp force = (numtyp)-2.0*gauss1[mtype].x*gauss1[mtype].y*rsq*
ucl_exp(-gauss1[mtype].y*rsq)*r2inv*factor_lj;
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
numtyp e=-(gauss1[mtype].x*ucl_exp(-gauss1[mtype].y*rsq) -
gauss1[mtype].w);
energy+=factor_lj*e;
}
if (vflag>0) {
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
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
}