[Pallas:MGPU] Add support for indexing untiled dimensions under WG semantic. #33973
+68
−5
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
…mantic.
Indexing is lowered to `lax.slice_p` + `lax.squeeze_p`. We add the missing rule for `lax.squeeze_p` under WG semantics. We lower `lax.squeeze_p` to `vector.shape_cast`.
So `x[1, 1]` where `x` is a JAX array will be lowered to:
```
%1 = vector.extract_strided_slice %0 {offsets = [1, 1, ...], sizes = [1, 1, ...]} : vector<NxMx...> to vector<1x1x...>
%2 = vector.shape_cast %2 : vector<1x1x...> to vector<...>
```
This gets simplified to:
```
%2 = vector.extract %0[1, 1] : vector<...> from vector<NxMx...>
```
So we need to add support for `vector.extract` to the Mosaic GPU dialect. We add the respective lowering and layout inference rules.
PiperOrigin-RevId: 844759978
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
[Pallas:MGPU] Add support for indexing untiled dimensions under WG semantic.
Indexing is lowered to
lax.slice_p+lax.squeeze_p. We add the missing rule forlax.squeeze_punder WG semantics. We lowerlax.squeeze_ptovector.shape_cast.So
x[1, 1]wherexis a JAX array will be lowered to:This gets simplified to:
So we need to add support for
vector.extractto the Mosaic GPU dialect. We add the respective lowering and layout inference rules.