Plumb Sparse Gaussian TileIntersection #401
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
…hecks Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
Swap out torch cumsum for CUB. Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
…r PR Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
…seTileIntersection to be more consistent with dispatchGaussianSparseRasterize* functions Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
swahtz
added
the
Gaussian Splatting
Signed-off-by: Jonathan Swartz <jonathan@jswartz.info>
harrism
approved these changes
Jan 13, 2026
Contributor
harrism
left a comment
harrism
left a comment
There was a problem hiding this comment.
This is awesome! Thanks for plumbing this in. Perhaps before merging change the title to not use "pipeline" since that confused me (I interpreted it to mean pipeline parallelism as in overlap multiple stages of an algorithm when processing independent inputs).
swahtz
changed the title
swahtz
commented
Jan 13, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When profiling sparse rasterization in functions like
sparseRenderTopContributingGaussianIds, I was noticing that the projection phase, specifically the gaussian tile intersection, dominated the runtime. However, these intersections were being computed for all tiles when we only eventually rasterize a small number of them. I went back and plumbeddispatchGaussianSparseTileIntersectionwhich had been written back in February of 2025 and this has provided a 7x speedup for the tile intersection cost and a 4x speedup tosparseRenderTopContributingGaussianIdsoverall.The most significant downstream change is that now
tileOffsetscan be either a tensor of dimension 3[C, numTilesH, numTilesW]in the usual dense rendering case or[active_tiles + 1]in the sparse case. This just necessitated changing some functions that relied ontileOffsets's sizes to infer any of those quantities to get them from elsewhere.