Reduction of BVH-node bandwidth with incremental traversal
Incremental encoding of Bounding Volume Hierarchies (BVH) enables coarse quantization of bounding volumes, significantly reducing their memory footprint. However, reducing the size of the BVH alone does not yield a comparable reduction in memory bandwidth in some embodiments. While the bounding volumes of the BVH nodes can be aggressively quantized, the size of the child node pointers remains a significant overhead. A two-level clustering method introduces a memory layout and node addressing scheme, which allows the reordering of BVH nodes to reduce their memory footprint in hardware ray tracing systems using reduced precision ray traversal.
1. A method comprising:
implementing reduced precision ray traversal for a bounding volume-node hierarchy in a graphics processor by using address clustering;
and
connecting two address clusters by adding a glue node between the two address clusters.
2. The method of claim 1 including generating low resolution child pointers between clustered nodes.
3. The method of claim 1 including performing cache aware node reordering within each cluster.
4. The method of claim 1 including only loading glue nodes when traversing between the two address clusters.
5. The method of claim 1 including merging child nodes in a cluster until the cluster is full.
6. The method of claim 1 including moving full clusters that match a cache line size to the beginning of an address cluster.
7. The method of claim 1 including storing cache clusters in a multi-level ordered depth-first layout.
8. The method of claim 1 including processing nodes using a traversal unit that traverses one ray per thread and storing a traversal state for the ray.
9. One or more non-transitory computer readable media storing instructions to perform a sequence comprising:
implementing reduced precision ray traversal for a bounding volume-node hierarchy in a graphics processor by using address clustering;
and
connecting two address clusters by adding a glue node between the two address clusters.
10. The media of claim 9 , further storing instructions to perform a sequence including generating low resolution child pointers between clustered nodes.
11. The media of claim 9 , further storing instructions to perform a sequence including performing cache aware node reordering within each cluster.
12. The media of claim 9 , further storing instructions to perform a sequence including only loading glue nodes when traversing between the two address clusters.
13. The media of claim 9 , further storing instructions to perform a sequence including merging child nodes in a cluster until the cluster is full.
14. The media of claim 9 , further storing instructions to perform a sequence including moving full clusters that match a cache line size to the beginning of an address cluster.
15. The media of claim 9 , further storing instructions to perform a sequence including storing cache clusters in a multi-level ordered depth-first layout.
16. The media of claim 9 , further storing instructions to perform a sequence including processing nodes using a traversal unit that traverses one ray per thread and storing a traversal state for the ray.
17. An apparatus comprising:
a processor to implement reduced precision ray traversal for a bounding volume-node hierarchy, by using address clustering, connect two address clusters by adding a glue node between the two address clusters; and
a memory coupled to said processor.
18. The apparatus of claim 17 , said processor to generate low resolution child pointers between clustered nodes.
19. The apparatus of claim 17 , said processor to perform cache aware node reordering within each cluster.
20. The apparatus of claim 17 , said processor to only load glue nodes when traversing between the two address clusters.
21. The apparatus of claim 17 , said processor to merge child nodes in a cluster until the cluster is full.
22. The apparatus of claim 17 , said processor to move full clusters that match a cache line size to the beginning of an address cluster.
23. The apparatus of claim 17 , said processor to store cache clusters in a multi-level ordered depth-first layout.
24. The apparatus of claim 17 , said processor to process nodes using a traversal unit that traverses one ray per thread and storing a traversal state for the ray.