build on cpu, ship to gpu for query
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 89
- Forks
- 24
- PR merge metrics
- No merged PRs in 30d
Description
Hello,
On my system where the GPU is shared among several processes, it might make more sense to build the BVH on the CPU and then ship them to the GPU for the NN query, which is what is time sensitive and needs to happen on the GPU. This is what I was doing when I was using cukd instead of cubql. Is this possible? In cukd the tree was a compact struct that could be memcpy-ed, so I guess that's the question here: is the bvh memory representation simple enough that it can be memcpy-ed to the GPU and be usable there?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reviewing cuBQL's BVH build and query interfaces, then compare the issue's cukd example of a compact, memcpy-able tree. Determine whether cuBQL can build on the CPU and transfer a usable BVH to the GPU for queries; done means the feasibility and required API or representation changes are established.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- computer-graphics, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100