[XLA:GPU] Add num_blocks to TiledHloComputation. #70897
Merged
+68
−19
Conversation
This file contains 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
6ffc08c
to
e57eaf4
Compare
The original plan was to use `block_id_to_tile_offsets_indexing` to get this information, but turned out that it's too expensive to compute IndexingMap for each tiled instruction. This is a step to minimize the places where we compute the indexing. The only case when number of blocks per tile is different is if the have a concat in the fusion. This case would have been supported by ``block_id_to_tile_offsets_indexing`, but now we need a new plan. Since `SymbolicTileAnalysis` also doesn't support `concat` yet, it's not an immediate issue. PiperOrigin-RevId: 650230997
e57eaf4
to
c80b67a
Compare
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Outdated suggestions cannot be applied.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.
[XLA:GPU] Add num_blocks to TiledHloComputation.
The original plan was to use
block_id_to_tile_offsets_indexingto get this information, but turned out that it's too expensive to compute IndexingMap for each tiled instruction. This is a step to minimize the places where we compute the indexing.The only case when number of blocks per tile is different is if the have a concat in the fusion. This case would have been supported by ``block_id_to_tile_offsets_indexing
, but now we need a new plan. SinceSymbolicTileAnalysis` also doesn't support `concat` yet, it's not an immediate issue.