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

Untested update of entire GPU library to use multiple threads per atom.

Browse Source
This commit is contained in:
W. Michael Brown
2011-04-14 17:25:55 -04:00
parent 2b37bf3d41
commit e27ca98bb8
30 changed files with 2135 additions and 784 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
lib/gpu/cmmc_long_gpu_memory.cpp
View File

@ -137,7 +137,8 @@ void CMML_GPU_MemoryT::loop(const bool _eflag, const bool _vflag) {
else
vflag=0;
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/BX));
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom)));
int ainum=this->ans->inum();
int anall=this->atom->nall();
@ -148,19 +149,21 @@ void CMML_GPU_MemoryT::loop(const bool _eflag, const bool _vflag) {
this->k_pair_fast.run(&this->atom->dev_x.begin(), &lj1.begin(),
&lj3.begin(), &sp_lj.begin(),
&this->nbor->dev_nbor.begin(),
&this->_nbor_data->begin(),
&this->ans->dev_ans.begin(),
&this->ans->dev_engv.begin(), &eflag, &vflag,
&ainum, &anall, &nbor_pitch,
&this->atom->dev_q.begin(), &_cut_coulsq,
&_qqrd2e, &_g_ewald);
&_qqrd2e, &_g_ewald, &this->_threads_per_atom);
} else {
this->k_pair.set_size(GX,BX);
this->k_pair.run(&this->atom->dev_x.begin(), &lj1.begin(), &lj3.begin(),
&_lj_types, &sp_lj.begin(), &this->nbor->dev_nbor.begin(),
&this->ans->dev_ans.begin(),
&this->_nbor_data->begin(), &this->ans->dev_ans.begin(),
&this->ans->dev_engv.begin(), &eflag, &vflag, &ainum,
&anall, &nbor_pitch, &this->atom->dev_q.begin(),
&_cut_coulsq, &_qqrd2e, &_g_ewald);
&_cut_coulsq, &_qqrd2e, &_g_ewald,
&this->_threads_per_atom);
}
this->time_pair.stop();
}