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

Merge pull request #2095 from stanmoore1/kk_snap_beta 

Fix redundant computation in Kokkos Pair SNAP on GPUs
This commit is contained in:
Axel Kohlmeyer 2020-05-20 15:30:17 -04:00 committed by GitHub
parent 58ca1dcf8e c67ca88b04
commit 5281dea651
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 4 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/KOKKOS/pair_snap_kokkos.h
View File

@ -112,7 +112,7 @@ public:
void operator() (TagPairSNAPComputeDeidrjCPU,const typename Kokkos::TeamPolicy<DeviceType, TagPairSNAPComputeDeidrjCPU>::member_type& team) const;
KOKKOS_INLINE_FUNCTION
void operator() (TagPairSNAPBeta,const typename Kokkos::TeamPolicy<DeviceType, TagPairSNAPBeta>::member_type& team) const;
void operator() (TagPairSNAPBeta,const int& ii) const;
template<int NEIGHFLAG>
KOKKOS_INLINE_FUNCTION

13
src/KOKKOS/pair_snap_kokkos_impl.h
View File

@ -284,13 +284,8 @@ void PairSNAPKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
}
//Compute beta = dE_i/dB_i for all i in list
{
int vector_length = vector_length_default;
int team_size = team_size_default;
check_team_size_for<TagPairSNAPBeta>(chunk_size,team_size,vector_length);
typename Kokkos::TeamPolicy<DeviceType,TagPairSNAPBeta> policy_beta(chunk_size,team_size,vector_length);
Kokkos::parallel_for("ComputeBeta",policy_beta,*this);
}
typename Kokkos::RangePolicy<DeviceType,TagPairSNAPBeta> policy_beta(0,chunk_size);
Kokkos::parallel_for("ComputeBeta",policy_beta,*this);
//ZeroYi
{
@ -424,10 +419,8 @@ void PairSNAPKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
template<class DeviceType>
KOKKOS_INLINE_FUNCTION
void PairSNAPKokkos<DeviceType>::operator() (TagPairSNAPBeta,const typename Kokkos::TeamPolicy<DeviceType,TagPairSNAPBeta>::member_type& team) const {
void PairSNAPKokkos<DeviceType>::operator() (TagPairSNAPBeta,const int& ii) const {
// TODO: use RangePolicy instead, or thread over ncoeff?
int ii = team.league_rank();
const int i = d_ilist[ii + chunk_offset];
const int itype = type[i];
const int ielem = d_map[itype];