warp.tile_bvh_query_next

warp.tile_bvh_query_next(

query: BvhQueryTiled,

) → Tile[int32, tuple[int]]

• Kernel: true
• Python: false
• Differentiable: false

Advance a thread-block parallel BVH query and return the next batch of results as a tile.

Every thread in the block must call this on each loop iteration. Guard the traversal loop with tile_query_valid() and test each returned index for >= 0.

Parameters:

query – The query to advance, from tile_bvh_query_aabb() or tile_bvh_query_ray()

Returns:

A register tile of shape (block_dim,) and dtype int holding one result index per thread: the index of an overlapping bound into the lowers/uppers arrays passed to warp.Bvh, or a negative value when that thread has no result.

See tile_bvh_query_aabb() for a usage example.