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

Feb2021 GPU Package Update - GPU Package Files

Browse Source
This commit is contained in:
Michael Brown
2021-02-15 08:20:50 -08:00
parent 16004e8f45
commit e7e2d2323b
345 changed files with 13424 additions and 7708 deletions
Download Patch File Download Diff File Expand all files Collapse all files

140
lib/gpu/lal_coul_long.cu
View File

@ -29,100 +29,6 @@ _texture( q_tex,int2);
#define q_tex q_
#endif
#if (ARCH < 300)
#define store_answers_lq(f, e_coul, virial, ii, inum, tid, \
t_per_atom, offset, eflag, vflag, ans, engv) \
if (t_per_atom>1) { \
__local acctyp red_acc[6][BLOCK_PAIR]; \
\
red_acc[0][tid]=f.x; \
red_acc[1][tid]=f.y; \
red_acc[2][tid]=f.z; \
red_acc[3][tid]=e_coul; \
\
for (unsigned int s=t_per_atom/2; s>0; s>>=1) { \
if (offset < s) { \
for (int r=0; r<4; r++) \
red_acc[r][tid] += red_acc[r][tid+s]; \
} \
} \
\
f.x=red_acc[0][tid]; \
f.y=red_acc[1][tid]; \
f.z=red_acc[2][tid]; \
e_coul=red_acc[3][tid]; \
\
if (vflag>0) { \
for (int r=0; r<6; r++) \
red_acc[r][tid]=virial[r]; \
\
for (unsigned int s=t_per_atom/2; s>0; s>>=1) { \
if (offset < s) { \
for (int r=0; r<6; r++) \
red_acc[r][tid] += red_acc[r][tid+s]; \
} \
} \
\
for (int r=0; r<6; r++) \
virial[r]=red_acc[r][tid]; \
} \
} \
\
if (offset==0) { \
__global acctyp *ap1=engv+ii; \
if (eflag>0) { \
*ap1=(acctyp)0; \
ap1+=inum; \
*ap1=e_coul*(acctyp)0.5; \
ap1+=inum; \
} \
if (vflag>0) { \
for (int i=0; i<6; i++) { \
*ap1=virial[i]*(acctyp)0.5; \
ap1+=inum; \
} \
} \
ans[ii]=f; \
}
#else
#define store_answers_lq(f, e_coul, virial, ii, inum, tid, \
t_per_atom, offset, eflag, vflag, ans, engv) \
if (t_per_atom>1) { \
for (unsigned int s=t_per_atom/2; s>0; s>>=1) { \
f.x += shfl_xor(f.x, s, t_per_atom); \
f.y += shfl_xor(f.y, s, t_per_atom); \
f.z += shfl_xor(f.z, s, t_per_atom); \
e_coul += shfl_xor(e_coul, s, t_per_atom); \
} \
if (vflag>0) { \
for (unsigned int s=t_per_atom/2; s>0; s>>=1) { \
for (int r=0; r<6; r++) \
virial[r] += shfl_xor(virial[r], s, t_per_atom); \
} \
} \
} \
if (offset==0) { \
__global acctyp *ap1=engv+ii; \
if (eflag>0) { \
*ap1=(acctyp)0; \
ap1+=inum; \
*ap1=e_coul*(acctyp)0.5; \
ap1+=inum; \
} \
if (vflag>0) { \
for (int i=0; i<6; i++) { \
*ap1=virial[i]*(acctyp)0.5; \
ap1+=inum; \
} \
} \
ans[ii]=f; \
}
#endif
__kernel void k_coul_long(const __global numtyp4 *restrict x_,
const __global numtyp *restrict scale,
const int lj_types,
@ -140,22 +46,25 @@ __kernel void k_coul_long(const __global numtyp4 *restrict x_,
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];
acctyp e_coul=(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;
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;
__local int n_stride;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
@ -197,10 +106,10 @@ __kernel void k_coul_long(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
e_coul += prefactor*(_erfc-factor_coul);
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -211,9 +120,11 @@ __kernel void k_coul_long(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers_lq(f,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
} // 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_fast(const __global numtyp4 *restrict x_,
@ -233,24 +144,27 @@ __kernel void k_coul_long_fast(const __global numtyp4 *restrict x_,
__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];
acctyp e_coul=(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;
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;
__local int n_stride;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
@ -292,10 +206,10 @@ __kernel void k_coul_long_fast(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
e_coul += prefactor*(_erfc-factor_coul);
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -306,8 +220,10 @@ __kernel void k_coul_long_fast(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers_lq(f,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
} // 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);
}