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

Added timing for the induced dipole spreading part, computed the block size to ensure all the CUs are occupied by the fphi_uind and fphi_mpole kernels

Browse Source
This commit is contained in:
Trung Nguyen
2022-10-06 15:03:58 -05:00
parent 009ed36301
commit 6b9e83fe20
10 changed files with 106 additions and 87 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
lib/gpu/lal_hippo.cpp
View File

@ -619,9 +619,14 @@ int HippoT::polar_real(const int eflag, const int vflag) {
int nbor_pitch=this->nbor->nbor_pitch();
// Compute the block size and grid size to keep all cores busy
const int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom)));
const int max_cus = this->device->max_cus();
int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(ainum)/(BX/this->_threads_per_atom)));
while (GX < max_cus) {
BX /= 2;
GX=static_cast<int>(ceil(static_cast<double>(ainum)/(BX/this->_threads_per_atom)));
}
this->time_pair.start();
// Build the short neighbor list if not done yet