Skip to content

Minor optimization in ExpandHalfToFull #1264

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
aprokop wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Minor optimization in ExpandHalfToFull #1264

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cd34718
Minor optimization in ExpandHalfToFull
aprokop Jun 11, 2025
0613c5f
Change the order of stored neighbors to improve traversal
aprokop Nov 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 18 additions & 9 deletions src/spatial/detail/ArborX_ExpandHalfToFull.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,10 +34,15 @@ void expandHalfToFull(ExecutionSpace const &space, Offsets &offsets,
Kokkos::parallel_for(
"ArborX::Experimental::HalfToFull::count",
Kokkos::RangePolicy(space, 0, n), KOKKOS_LAMBDA(int i) {
for (int j = offsets_orig(i); j < offsets_orig(i + 1); ++j)
auto const start = offsets_orig(i);
auto const end = offsets_orig(i + 1);
if (start == end)
return;

Kokkos::atomic_add(&offsets(i), end - start);
for (auto j = start; j < end; ++j)
{
int const k = indices_orig(j);
Kokkos::atomic_inc(&offsets(i));
auto const k = indices_orig(j);
Comment thread
aprokop marked this conversation as resolved.
Kokkos::atomic_inc(&offsets(k));
}
});
Expand All @@ -55,13 +60,17 @@ void expandHalfToFull(ExecutionSpace const &space, Offsets &offsets,
typename Kokkos::TeamPolicy<ExecutionSpace>::member_type const
&member) {
auto const i = member.league_rank();
auto const first = offsets_orig(i);
auto const last = offsets_orig(i + 1);
auto const start = offsets_orig(i);
auto const end = offsets_orig(i + 1);
if (start == end)
return;

auto const offset = offsets(i);
Kokkos::parallel_for(
Kokkos::TeamVectorRange(member, last - first), [&](int j) {
int const k = indices_orig(first + j);
indices(Kokkos::atomic_fetch_inc(&counts(i))) = k;
indices(Kokkos::atomic_fetch_inc(&counts(k))) = i;
Kokkos::TeamVectorRange(member, end - start), [&](int j) {
auto const k = indices_orig(start + j);
indices(offset + j) = k;
indices(Kokkos::atomic_dec_fetch(&counts(k + 1))) = i;
});
});
Kokkos::Profiling::popRegion();
Expand Down
15 changes: 7 additions & 8 deletions src/spatial/detail/ArborX_NeighborList.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,8 +81,8 @@ void findHalfNeighborList(ExecutionSpace const &space,
Kokkos::deep_copy(space, offsets, 0);
HalfTraversal(
space, bvh,
KOKKOS_LAMBDA(Value const &, Value const &value) {
Kokkos::atomic_inc(&offsets(value.index));
KOKKOS_LAMBDA(Value const &value1, Value const &) {
Kokkos::atomic_inc(&offsets(value1.index));
},
NeighborListPredicateGetter<Coordinate>{radius});
KokkosExt::exclusive_scan(space, offsets, offsets, 0);
Expand All @@ -98,7 +98,7 @@ void findHalfNeighborList(ExecutionSpace const &space,
HalfTraversal(
space, bvh,
KOKKOS_LAMBDA(Value const &value1, Value const &value2) {
indices(Kokkos::atomic_fetch_inc(&counts(value2.index))) = value1.index;
indices(Kokkos::atomic_fetch_inc(&counts(value1.index))) = value2.index;
},
Comment thread
aprokop marked this conversation as resolved.
NeighborListPredicateGetter<Coordinate>{radius});

Expand Down Expand Up @@ -159,7 +159,7 @@ void findFullNeighborList(ExecutionSpace const &space,
HalfTraversal(
space, bvh,
KOKKOS_LAMBDA(Value const &value1, Value const &value2) {
indices(Kokkos::atomic_fetch_inc(&counts(value2.index))) = value1.index;
indices(Kokkos::atomic_fetch_inc(&counts(value1.index))) = value2.index;
},
NeighborListPredicateGetter<Coordinate>{radius});

Expand All @@ -174,11 +174,10 @@ void findFullNeighborList(ExecutionSpace const &space,
typename Kokkos::TeamPolicy<ExecutionSpace>::member_type const
&member) {
auto const i = member.league_rank();
auto const first = offsets(i);
auto const last = counts_copy(i);
Kokkos::parallel_for(
Kokkos::TeamVectorRange(member, last - first), [&](int j) {
int const k = indices(first + j);
Kokkos::TeamVectorRange(member, offsets(i), counts_copy(i)),
[&](int j) {
int const k = indices(j);
indices(Kokkos::atomic_fetch_inc(&counts(k))) = i;
});
});
Expand Down
Loading