jianmu
/
lammps
forked from lijiext/lammps
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix CUDA issue in pair_exp6_rx_kokkos

This commit is contained in:
Stan Moore 2018-05-07 15:51:15 -06:00
parent 624173d25a
commit cde6e9b7c4
1 changed files with 12 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/KOKKOS/pair_exp6_rx_kokkos.cpp
View File

@ -813,10 +813,13 @@ void PairExp6rxKokkos<DeviceType>::operator()(TagPairExp6rxComputeNoAtomics<NEIG
return;
}
int tid = 0;
#ifndef KOKKOS_HAVE_CUDA
typedef Kokkos::Experimental::UniqueToken<
DeviceType, Kokkos::Experimental::UniqueTokenScope::Global> unique_token_type;
unique_token_type unique_token;
int tid = unique_token.acquire();
tid = unique_token.acquire();
#endif
int i,jj,jnum,itype,jtype;
double xtmp,ytmp,ztmp,delx,dely,delz,evdwl,evdwlOld,fpair;
@ -1153,7 +1156,9 @@ void PairExp6rxKokkos<DeviceType>::operator()(TagPairExp6rxComputeNoAtomics<NEIG
t_uCG(tid,i) += uCG_i;
t_uCGnew(tid,i) += uCGnew_i;
#ifndef KOKKOS_HAVE_CUDA
unique_token.release(tid);
#endif
}
// Experimental thread-safe approach using duplicated data instead of atomics and
@ -1183,10 +1188,13 @@ void PairExp6rxKokkos<DeviceType>::vectorized_operator(const int &ii, EV_FLOAT&
Kokkos::View<E_FLOAT*, typename DAT::t_efloat_1d::array_layout,DeviceType,Kokkos::MemoryTraits<AtomicF<NEIGHFLAG>::value> > a_uCG = uCG;
Kokkos::View<E_FLOAT*, typename DAT::t_efloat_1d::array_layout,DeviceType,Kokkos::MemoryTraits<AtomicF<NEIGHFLAG>::value> > a_uCGnew = uCGnew;
int tid = 0;
#ifndef KOKKOS_HAVE_CUDA
typedef Kokkos::Experimental::UniqueToken<
DeviceType, Kokkos::Experimental::UniqueTokenScope::Global> unique_token_type;
unique_token_type unique_token;
int tid = unique_token.acquire();
tid = unique_token.acquire();
#endif
const int nRep = 12;
const double shift = 1.05;
@ -1615,7 +1623,9 @@ void PairExp6rxKokkos<DeviceType>::vectorized_operator(const int &ii, EV_FLOAT&
t_uCGnew(tid,i) += uCGnew_i;
}
#ifndef KOKKOS_HAVE_CUDA
unique_token.release(tid);
#endif
}
template<class DeviceType>