Repository navigation
Conversation
Keep independent buffer layouts while an isolated compatible L0B pair uses one version clock and physical owner dependencies. Preserve the ordinary path for incompatible or nested owners. Validate retimed and unequal owner lengths, allocation reuse controls, and repeated device launches.
Denverjin
changed the title
Denverjin
changed the title
Denverjin
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Keep layouts attached to individual buffers created by T.view or T.reshape while their shared .data determines allocation and version ownership. Cube operands can use distinct MajorK/MajorMN layouts on one L0B ring, and UB views retain their own annotated address mappings.
Infer each view's layout independently and inherit missing layouts only when unambiguous. Normalize Cube padding per Buffer, match reshaped views to their own matrix layout, and check shared physical capacity after padding. Preserve typed pointers through Copy/GEMM lowering so each access uses the correct view.
Cube aliases require dense logical strides, zero base offset, equal element widths, and equal padded storage sizes. Views reinterpret existing storage without moving values; producers must write the representation expected by consumers. Use T.access_ptr(view[...], ...) when passing a view to an intrinsic.