Introduce AllGatherDynamicSliceShuffledOffsetSimplifier, a new HLO pass that collapse dynamic-slice(all-gather) with shuffled offset into collective-permute.
#97287
+492
−0
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
2cd582c to
5e9658e
Compare
…pass that collapse dynamic-slice(all-gather) with shuffled offset into collective-permute. The simplifier targets a specific pattern: * **All-gather flattened-id Mode:** The `all-gather` must use `use_global_device_ids=true`. * **Partitioning:** The pass is designed for SPMD models with `num_partitions > 1` and `num_replicas = 1`. The slice offset is indexed by `partition-id()`. * **Sharding and Slicing on same dimension:**: The `dynamic-slice` must also slice along that same dimension, matching the `all-gather` operand's dimension size. * **Offset Traversal Calculation:** The pass can trace the offset lookup through `bitcast`, `reshape`, `copy`, and `convert` instructions. PiperOrigin-RevId: 785634123
5e9658e to
31b031c
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Outdated suggestions cannot be applied.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.
Introduce
AllGatherDynamicSliceShuffledOffsetSimplifier, a new HLO pass that collapse dynamic-slice(all-gather) with shuffled offset into collective-permute.The simplifier targets a specific pattern:
all-gathermust useuse_global_device_ids=true.num_partitions > 1andnum_replicas = 1. The slice offset is indexed bypartition-id().dynamic-slicemust also slice along that same dimension, matching theall-gatheroperand's dimension size.bitcast,reshape,copy, andconvertinstructions.