Improve prefetch granularity for rasterization kernels#665
Merged
matthewdcong merged 2 commits intoJun 8, 2026
Conversation
matthewdcong
force-pushed
the
improve_rasterization_prefetch
branch
from
1b756fb to
979d58a
Compare
swahtz
reviewed
Jun 8, 2026
swahtz
requested changes
Jun 8, 2026
swahtz
left a comment
swahtz
left a comment
Contributor
There was a problem hiding this comment.
Overall it looks good, I think we should just add a few gtests for the logic in Prefetch's appendPerTileImageRanges to make sure we don't have any silent regressions in logic before we merge this.
Signed-off-by: Matthew Cong <mcong@nvidia.com>
Signed-off-by: Matthew Cong <mcong@nvidia.com>
matthewdcong
force-pushed
the
improve_rasterization_prefetch
branch
from
d754157 to
6342627
Compare
swahtz
approved these changes
Jun 8, 2026
swahtz
left a comment
swahtz
left a comment
Contributor
There was a problem hiding this comment.
Looks good to me, thanks for adding those tests!
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Prior to this PR, we prefetched image-space tensors with camera granularity. That is, if an image in a camera was rendered across multiple GPUs, each GPU would prefetch the entire image. This is overly conservative and increases the amount of bandwidth required.
This PR refines that prefetching to have better granularity by prefetching on the tiles rendered by each GPU. This reduces the amount of cross-device traffic and leads to an approximately 5% speedup on 2x RTX 3090s for training.