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

169
lib/gpu/lal_table.cu
View File

@ -58,24 +58,27 @@ __kernel void k_table(const __global numtyp4 *restrict x_,
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
int tlm1 = tablength - 1;
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);
@ -112,13 +115,13 @@ __kernel void k_table(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1)
e = coeff3[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -129,9 +132,9 @@ __kernel void k_table(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
__kernel void k_table_fast(const __global numtyp4 *restrict x_,
@ -153,18 +156,22 @@ __kernel void k_table_fast(const __global numtyp4 *restrict x_,
__local numtyp cutsq[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
if (tid<4)
sp_lj[tid]=sp_lj_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
cutsq[tid]=cutsq_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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
@ -173,7 +180,6 @@ __kernel void k_table_fast(const __global numtyp4 *restrict x_,
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);
@ -211,13 +217,13 @@ __kernel void k_table_fast(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1)
e = coeff3[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -228,9 +234,9 @@ __kernel void k_table_fast(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
/// ---------------- LINEAR -------------------------------------------------
@ -254,24 +260,27 @@ __kernel void k_table_linear(const __global numtyp4 *restrict x_,
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
int tlm1 = tablength - 1;
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);
@ -312,13 +321,13 @@ __kernel void k_table_linear(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1)
e = coeff3[idx].y + fraction*coeff4[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -329,9 +338,9 @@ __kernel void k_table_linear(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
__kernel void k_table_linear_fast(const __global numtyp4 *restrict x_,
@ -353,18 +362,22 @@ __kernel void k_table_linear_fast(const __global numtyp4 *restrict x_,
__local numtyp cutsq[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
if (tid<4)
sp_lj[tid]=sp_lj_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
cutsq[tid]=cutsq_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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
@ -373,7 +386,6 @@ __kernel void k_table_linear_fast(const __global numtyp4 *restrict x_,
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);
@ -415,13 +427,13 @@ __kernel void k_table_linear_fast(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1)
e = coeff3[idx].y + fraction*coeff4[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -432,9 +444,9 @@ __kernel void k_table_linear_fast(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
/// ---------------- SPLINE -------------------------------------------------
@ -458,24 +470,27 @@ __kernel void k_table_spline(const __global numtyp4 *restrict x_,
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
int tlm1 = tablength - 1;
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);
@ -520,7 +535,7 @@ __kernel void k_table_spline(const __global numtyp4 *restrict x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1) {
e = a * coeff3[idx].y + b * coeff3[idx+1].y +
@ -529,7 +544,7 @@ __kernel void k_table_spline(const __global numtyp4 *restrict x_,
}
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -540,9 +555,9 @@ __kernel void k_table_spline(const __global numtyp4 *restrict x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
__kernel void k_table_spline_fast(const __global numtyp4 *x_,
@ -564,19 +579,22 @@ __kernel void k_table_spline_fast(const __global numtyp4 *x_,
__local numtyp cutsq[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
if (tid<4)
sp_lj[tid]=sp_lj_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
cutsq[tid]=cutsq_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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
int tlm1 = tablength - 1;
@ -584,7 +602,6 @@ __kernel void k_table_spline_fast(const __global numtyp4 *x_,
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);
@ -630,7 +647,7 @@ __kernel void k_table_spline_fast(const __global numtyp4 *x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable < tlm1) {
e = a * coeff3[idx].y + b * coeff3[idx+1].y +
@ -639,7 +656,7 @@ __kernel void k_table_spline_fast(const __global numtyp4 *x_,
}
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -650,9 +667,9 @@ __kernel void k_table_spline_fast(const __global numtyp4 *x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
/// ---------------- BITMAP -------------------------------------------------
@ -678,24 +695,27 @@ __kernel void k_table_bitmap(const __global numtyp4 *x_,
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
int tlm1 = tablength - 1;
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);
@ -739,13 +759,13 @@ __kernel void k_table_bitmap(const __global numtyp4 *x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable <= tlm1)
e = coeff3[idx].y + fraction*coeff4[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -756,9 +776,9 @@ __kernel void k_table_bitmap(const __global numtyp4 *x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}
__kernel void k_table_bitmap_fast(const __global numtyp4 *x_,
@ -782,18 +802,22 @@ __kernel void k_table_bitmap_fast(const __global numtyp4 *x_,
__local numtyp cutsq[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_lj[4];
int n_stride;
local_allocate_store_pair();
if (tid<4)
sp_lj[tid]=sp_lj_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
cutsq[tid]=cutsq_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;
acctyp energy, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
@ -802,7 +826,6 @@ __kernel void k_table_bitmap_fast(const __global numtyp4 *x_,
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);
@ -847,13 +870,13 @@ __kernel void k_table_bitmap_fast(const __global numtyp4 *x_,
f.y+=dely*force;
f.z+=delz*force;
if (eflag>0) {
if (EVFLAG && eflag) {
numtyp e = (numtyp)0.0;
if (itable <= tlm1)
e = coeff3[idx].y + fraction*coeff4[idx].y;
energy+=factor_lj*e;
}
if (vflag>0) {
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
@ -864,7 +887,7 @@ __kernel void k_table_bitmap_fast(const __global numtyp4 *x_,
}
} // for nbor
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
} // if ii
store_answers(f,energy,virial,ii,inum,tid,t_per_atom,offset,eflag,vflag,
ans,engv);
}