[MLIR][AMDGPU] Use Attr for resetOffset + boundsCheck in RawBufferCastOp #149939
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
In order to access and modify resetOffset and boundsCheck of RawBufferCastOp in pythonic binding, we will have to use Attrs instead of Property. This is because we do not have python binding support for property yet. We should move back to property once we add pythonic binding support for it. Signed-off-by: Stanley Winata <[email protected]>
raikonenfnu
requested review from
ftynse,
makslevental and
stellaraccident
as code owners
Member
|
@llvm/pr-subscribers-mlir-gpu @llvm/pr-subscribers-backend-amdgpu Author: Stanley Winata (raikonenfnu) ChangesIn order to access and modify resetOffset and boundsCheck of RawBufferCastOp in pythonic binding, we will have to use Attrs instead of Property. This is because we do not have python binding support for property yet. We should move back to property once we add pythonic binding support for it. Full diff: https://github.com/llvm/llvm-project/pull/149939.diff 2 Files Affected:
diff --git a/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td b/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
index 80959ffbaf426..e2ca1ba9ae5e6 100644
--- a/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
+++ b/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
@@ -237,8 +237,8 @@ def AMDGPU_FatRawBufferCastOp :
Arguments<(ins AnyMemRef:$source,
Optional<I32>:$validBytes,
Optional<I<14>>:$cacheSwizzleStride,
- DefaultValuedProp<BoolProp, "true">:$boundsCheck,
- UnitProp:$resetOffset)>,
+ DefaultValuedAttr<BoolAttr, "true">:$boundsCheck,
+ UnitAttr:$resetOffset)>,
Results<(outs AnyMemRef:$result)> {
let summary = "Create a raw buffer fat pointer that matches `memref`";
let description = [{
diff --git a/mlir/test/python/dialects/amdgpu.py b/mlir/test/python/dialects/amdgpu.py
index c8039d494cf81..090888a94d667 100644
--- a/mlir/test/python/dialects/amdgpu.py
+++ b/mlir/test/python/dialects/amdgpu.py
@@ -2,7 +2,7 @@
# This is just a smoke test that the dialect is functional.
from mlir.ir import *
-from mlir.dialects import amdgpu, arith, memref
+from mlir.dialects import amdgpu, func
def constructAndPrintInModule(f):
@@ -20,3 +20,28 @@ def constructAndPrintInModule(f):
def testSmoke():
# CHECK: amdgpu.lds_barrier
amdgpu.LDSBarrierOp()
+
+
+# CHECK-LABEL: testFatRawBufferCastOpParams
+@constructAndPrintInModule
+def testFatRawBufferCastOpParams():
+ memref_type = MemRefType.get(
+ [ShapedType.get_dynamic_size(), ShapedType.get_dynamic_size()],
+ F32Type.get(),
+ )
+ f = func.FuncOp(
+ "test_raw_buffer_cast_params", ([memref_type], [])
+ )
+ with InsertionPoint(f.add_entry_block()):
+ block_args = f.arguments
+ amdgpu.FatRawBufferCastOp(block_args[0])
+ amdgpu.FatRawBufferCastOp(block_args[0], resetOffset=True)
+ amdgpu.FatRawBufferCastOp(block_args[0], boundsCheck=False)
+ amdgpu.FatRawBufferCastOp(block_args[0], boundsCheck=False, resetOffset=True)
+ func.ReturnOp([])
+
+ #CHECK: func.func @test_raw_buffer_cast_params(%[[ARG0:.+]]: memref<?x?xf32>) {
+ #CHECK: amdgpu.fat_raw_buffer_cast %[[ARG0]] : memref<?x?xf32> to memref<?x?xf32, #amdgpu.address_space<fat_raw_buffer>>
+ #CHECK-NEXT: amdgpu.fat_raw_buffer_cast %[[ARG0]] resetOffset : memref<?x?xf32> to memref<?x?xf32, #amdgpu.address_space<fat_raw_buffer>>
+ #CHECK-NEXT: amdgpu.fat_raw_buffer_cast %[[ARG0]] boundsCheck(false) : memref<?x?xf32> to memref<?x?xf32, #amdgpu.address_space<fat_raw_buffer>>
+ #CHECK-NEXT: amdgpu.fat_raw_buffer_cast %[[ARG0]] boundsCheck(false) resetOffset : memref<?x?xf32> to memref<?x?xf32, #amdgpu.address_space<fat_raw_buffer>>
|
|
✅ With the latest revision this PR passed the Python code formatter. |
Hardcode84
reviewed
Jul 21, 2025
| DefaultValuedProp<BoolProp, "true">:$boundsCheck, | ||
| UnitProp:$resetOffset)>, | ||
| DefaultValuedAttr<BoolAttr, "true">:$boundsCheck, | ||
| UnitAttr:$resetOffset)>, |
Contributor
There was a problem hiding this comment.
Pls, add a TODO somewhere nearby
Member
Author
There was a problem hiding this comment.
Added a TODO, but not sure this is best spot, let me know if you can think of a better spot :)
Member
Author
|
@makslevental, the man of the hour! Would be nice to kick off a chat wrt enabling prop in pythonic binding! |
Signed-off-by: Stanley Winata <[email protected]>
Hardcode84
approved these changes
Jul 21, 2025
mahesh-attarde
pushed a commit
to mahesh-attarde/llvm-project
that referenced
this pull request
Jul 28, 2025
…tOp (llvm#149939) In order to access and modify resetOffset and boundsCheck of RawBufferCastOp in pythonic binding, we will have to use Attrs instead of Property. This is because we do not have python binding support for property yet. We should move back to property once we add pythonic binding support for it. --------- Signed-off-by: Stanley Winata <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
In order to access and modify resetOffset and boundsCheck of RawBufferCastOp in pythonic binding, we will have to use Attrs instead of Property. This is because we do not have python binding support for property yet. We should move back to property once we add pythonic binding support for it.