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

Fixed bugs with zero local atoms (similar to what has been done to PPPM interp)

Browse Source
This commit is contained in:
Trung Nguyen
2021-09-16 17:27:44 -05:00
parent 98c1a0178c
commit c0b967054e
2 changed files with 23 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
lib/gpu/lal_amoeba.cpp
View File

@ -128,14 +128,18 @@ double AmoebaT::host_memory_usage() const {
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
template <class numtyp, class acctyp> template <class numtyp, class acctyp>
int AmoebaT::udirect2b(const int eflag, const int vflag) { int AmoebaT::udirect2b(const int eflag, const int vflag) {
int ainum=this->ans->inum();
if (ainum == 0)
return 0;
int _nall=this->atom->nall(); int _nall=this->atom->nall();
int nbor_pitch=this->nbor->nbor_pitch(); int nbor_pitch=this->nbor->nbor_pitch();
int ainum=this->ans->inum();
// Compute the block size and grid size to keep all cores busy // Compute the block size and grid size to keep all cores busy
const int BX=this->block_size(); const int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/ int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom))); (BX/this->_threads_per_atom)));
this->time_pair.start();
// Build the short neighbor list if not done yet // Build the short neighbor list if not done yet
if (!this->short_nbor_avail) { if (!this->short_nbor_avail) {
@ -164,14 +168,17 @@ int AmoebaT::udirect2b(const int eflag, const int vflag) {
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
template <class numtyp, class acctyp> template <class numtyp, class acctyp>
int AmoebaT::umutual2b(const int eflag, const int vflag) { int AmoebaT::umutual2b(const int eflag, const int vflag) {
int ainum=this->ans->inum();
if (ainum == 0)
return 0;
int _nall=this->atom->nall();
int nbor_pitch=this->nbor->nbor_pitch();
// Compute the block size and grid size to keep all cores busy // Compute the block size and grid size to keep all cores busy
const int BX=this->block_size(); const int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/ int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom))); (BX/this->_threads_per_atom)));
int _nall=this->atom->nall();
int ainum=this->ans->inum();
int nbor_pitch=this->nbor->nbor_pitch();
this->time_pair.start(); this->time_pair.start();
// Build the short neighbor list if not done yet // Build the short neighbor list if not done yet
@ -200,14 +207,17 @@ int AmoebaT::umutual2b(const int eflag, const int vflag) {
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
template <class numtyp, class acctyp> template <class numtyp, class acctyp>
int AmoebaT::polar_real(const int eflag, const int vflag) { int AmoebaT::polar_real(const int eflag, const int vflag) {
int ainum=this->ans->inum();
if (ainum == 0)
return 0;
int _nall=this->atom->nall();
int nbor_pitch=this->nbor->nbor_pitch();
// Compute the block size and grid size to keep all cores busy // Compute the block size and grid size to keep all cores busy
const int BX=this->block_size(); const int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/ int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom))); (BX/this->_threads_per_atom)));
int _nall=this->atom->nall();
int ainum=this->ans->inum();
int nbor_pitch=this->nbor->nbor_pitch();
this->time_pair.start(); this->time_pair.start();
// Build the short neighbor list if not done yet // Build the short neighbor list if not done yet

6
lib/gpu/lal_pppm.cpp
View File

@ -342,13 +342,15 @@ void PPPMT::interp(const grdtyp qqrd2e_scale) {
vd_brick.update_device(true); vd_brick.update_device(true);
time_in.stop(); time_in.stop();
int ainum=this->ans->inum();
if (ainum==0)
return;
time_interp.start(); time_interp.start();
// Compute the block size and grid size to keep all cores busy // Compute the block size and grid size to keep all cores busy
int BX=this->block_size(); int BX=this->block_size();
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));
int ainum=this->ans->inum();
k_interp.set_size(GX,BX); k_interp.set_size(GX,BX);
k_interp.run(&atom->x, &atom->q, &ainum, &vd_brick, &d_rho_coeff, k_interp.run(&atom->x, &atom->q, &ainum, &vd_brick, &d_rho_coeff,
&_npts_x, &_npts_yx, &_brick_x, &_brick_y, &_brick_z, &_delxinv, &_npts_x, &_npts_yx, &_brick_x, &_brick_y, &_brick_z, &_delxinv,