[DT] Implement MaterializeInterfaceBindingEncoding with interface methods. #22467
+117
−190
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
…hods. Signed-off-by: hanhanW <[email protected]>
|
I'll delete |
There was a problem hiding this comment.
Pull Request Overview
This PR refactors the encoding materialization infrastructure to consolidate type conversion logic from MaterializeEncodingIntoPadding.cpp into the external models in GPUEncodingExternalModels.cpp. The main goal is to implement the convertType interface method directly in the GPU padding resolver attribute, eliminating the need for a separate custom type converter and pattern for padding-specific encoding.
- Moves
getPadLayoutandgetPaddedTypehelper functions fromMaterializeEncodingIntoPadding.cpptoGPUEncodingExternalModels.cpp - Implements
convertTypemethod inGPUPadEncodingLayoutMaterializerAttrto handle type conversion for padding encodings - Simplifies
MaterializeEncodingTypeConverterby removing padding-specific special cases - Updates
MaterializeInterfaceBindingEncodingpattern to use the unified type conversion approach
Reviewed Changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
GPUEncodingExternalModels.cpp |
Adds getPadLayout and getPaddedType helper functions and implements convertType method for GPUPadEncodingLayoutMaterializerAttr |
materialize_encoding_into_padding.mlir |
Updates test expectations to include dynamic dimensions in dispatch tensor types |
MaterializeEncodingPatterns.cpp |
Refactors MaterializeInterfaceBindingEncoding to use generic type converter interface and compute dynamic dimensions via getOffsetsSizesStrides |
MaterializeEncodingIntoPadding.cpp |
Removes custom type converter and pattern implementations that are now handled by the external model |
EncodingUtils.cpp |
Removes padding-specific special cases from type converter, delegating all type conversion to the layout attribute's convertType method |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
compiler/src/iree/compiler/Codegen/ExternalInterfaces/GPUEncodingExternalModels.cpp
Show resolved
Hide resolved
kuhar
reviewed
Oct 29, 2025
compiler/src/iree/compiler/Codegen/Common/MaterializeEncodingPatterns.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/Common/MaterializeEncodingPatterns.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/Common/MaterializeEncodingPatterns.cpp
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/Common/MaterializeEncodingPatterns.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/ExternalInterfaces/GPUEncodingExternalModels.cpp
Outdated
Show resolved
Hide resolved
compiler/src/iree/compiler/Codegen/ExternalInterfaces/GPUEncodingExternalModels.cpp
Outdated
Show resolved
Hide resolved
Signed-off-by: hanhanW <[email protected]>
Signed-off-by: hanhanW <[email protected]>
kuhar
approved these changes
Oct 30, 2025
| return IREE::TensorExt::DispatchTensorType::get( | ||
| dispatchTensorType.getAccess(), type); | ||
| }) | ||
| .Default([&](Type type) { return type; }); |
There was a problem hiding this comment.
Suggested change
| .Default([&](Type type) { return type; }); | |
| .Default([](Type type) { return type; }); |
There was a problem hiding this comment.
oh thanks, somehow I missed it.
| Type convertType(Attribute attr, Type type) const { | ||
| auto layoutAttr = cast<IREE::Encoding::LayoutResolverAttr>(attr); | ||
| return TypeSwitch<Type, Type>(type) | ||
| .Case([&](RankedTensorType type) { |
There was a problem hiding this comment.
Suggested change
| .Case([&](RankedTensorType type) { | |
| .Case([](RankedTensorType type) { |
Signed-off-by: hanhanW <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
The revision implements
LayoutMaterializerAttr::convertTypefor GPUPaddingResolverAttr; updates theMaterializeInterfaceBindingEncodingpattern to use interface methods to materialize the bindings.It also fixes a bug in
@dynamic_set_zero_pad_encoding_and_storetest that misses dynamic dimensions in binding ops.It is a step towards #20160