Switch spheres to new hyper-dimensional ones #1146
Conversation
| @@ -55,7 +55,7 @@ struct ArborX::AccessTraits<Spheres, ArborX::PredicatesTag> | |||
| static KOKKOS_FUNCTION auto get(Spheres const &d, std::size_t i) | |||
| { | |||
| return ArborX::intersects( | |||
| ArborX::Sphere{{{d.d_x[i], d.d_y[i], d.d_z[i]}}, d.d_r[i]}); | |||
| ArborX::Sphere{ArborX::Point{d.d_x[i], d.d_y[i], d.d_z[i]}, d.d_r[i]}); | |||
| @@ -110,8 +110,7 @@ struct approx_expand_by_radius<PointTag, Point> | |||
| { | |||
| constexpr int DIM = GeometryTraits::dimension_v<Box>; | |||
| using Coordinate = GeometryTraits::coordinate_type_t<Point>; | |||
| return ExperimentalHyperGeometry::Sphere<DIM, Coordinate>{ | |||
| Kokkos::bit_cast<::ArborX::Point<DIM, Coordinate>>(point), r}; | |||
| return Sphere{Kokkos::bit_cast<::ArborX::Point<DIM, Coordinate>>(point), r}; | |||
There was a problem hiding this comment.
Is the bit_cast still necessary here?
There was a problem hiding this comment.
Yes. This geometries may come from users. We do an implicit undocumented cast here as we don't have a well defined concept of a point. It's ok for now but will need to change in the future.
| auto const &hyper_point = | ||
| reinterpret_cast<::ArborX::Point<dim, Coordinate> const &>(pair.value); |
There was a problem hiding this comment.
Do you still need the reinterpret_cast?
There was a problem hiding this comment.
Could be replaced by bit_cast but the issue is same as above: user data.
| auto const &hyper_point = | ||
| reinterpret_cast<::ArborX::Point<dim, Coordinate> const &>(point); |
| auto const &hyper_point = | ||
| reinterpret_cast<::ArborX::Point<dim> const &>(pair.value); |
| auto const &hyper_point = | ||
| reinterpret_cast<::ArborX::Point<dim> const &>(point); |
|
So, all the comments are about I think this is a larger discussion and is outside of the scope of this PR. |
|
@masterleinad I have a patch to address the conversion routines. Will do a separate PR after this one is merged. |
No description provided.