Repository navigation
Reduce memory pressure for Gaussian tile intersections - #839
Merged
Merged
Conversation
Scan one count per Gaussian block and reconstruct per-Gaussian offsets during emission on CUDA and multi-GPU paths. Signed-off-by: Matthew Cong <mcong@nvidia.com>
swahtz
reviewed
Oct 30, 2026
swahtz
reviewed
Oct 30, 2026
swahtz
approved these changes
Oct 30, 2026
swahtz
left a comment
swahtz
left a comment
Contributor
There was a problem hiding this comment.
Looks good. I saw two things worth addressing that are not correctness issues but might help slightly with performance or reduce likelihood of a future bug.
Signed-off-by: Matthew Cong <mcong@nvidia.com>
matthewdcong
force-pushed
the
block_gaussian_offsets
branch
from
October 30, 2026 15:41
31e79a8 to
48c1a18
Compare
matthewdcong
enabled auto-merge (squash)
October 30, 2026 16:02
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Tile intersection currently stores and scans a count per projected Gaussian, with a separate buffer on each GPU in mGPU. This PR reduces these buffers to one count per block and reconstructs per-Gaussian offsets using a block-local scan during intersection emission. This reduces temporary storage and global scan size for both CUDA and mGPU while providing identical results. The reduced global memory reads also speed up the kernel by 32% reducing single GPU training time end to end by 2-3%.