[mlir][xegpu] Support boundary checks only for block instructions #119380
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
Constrains Vector lowering to apply boundary checks only to data transfers operating on block shapes. This further aligns lowering with the current Xe instructions' restrictions.
Member
|
@llvm/pr-subscribers-mlir-gpu Author: Adam Siemieniuk (adam-smnk) ChangesConstrains Vector lowering to apply boundary checks only to data transfers operating on block shapes. This further aligns lowering with the current Xe instructions' restrictions. Full diff: https://github.com/llvm/llvm-project/pull/119380.diff 5 Files Affected:
diff --git a/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp b/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
index 215e1b1b874520..1232d8795d4dcf 100644
--- a/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
+++ b/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
@@ -82,6 +82,10 @@ static LogicalResult transferPreconditions(PatternRewriter &rewriter,
xferOp, "Buffer must be contiguous in the innermost dimension");
unsigned vecRank = vecTy.getRank();
+ if (xferOp.hasOutOfBoundsDim() && vecRank < 2)
+ return rewriter.notifyMatchFailure(
+ xferOp, "Boundary check is available only for block instructions.");
+
AffineMap map = xferOp.getPermutationMap();
if (!map.isProjectedPermutation(/*allowZeroInResults=*/false))
return rewriter.notifyMatchFailure(xferOp, "Unsupported permutation map");
@@ -255,9 +259,12 @@ struct LoadLowering : public OpRewritePattern<vector::LoadOp> {
if (failed(storeLoadPreconditions(rewriter, loadOp, vecTy)))
return failure();
+ // Boundary check is available only for block instructions.
+ bool boundaryCheck = vecTy.getRank() > 1;
+
auto descType = xegpu::TensorDescType::get(
vecTy.getShape(), vecTy.getElementType(), /*array_length=*/1,
- /*boundary_check=*/true, xegpu::MemorySpace::Global);
+ boundaryCheck, xegpu::MemorySpace::Global);
xegpu::CreateNdDescOp ndDesc = createNdDescriptor(
rewriter, loc, descType, loadOp.getBase(), loadOp.getIndices());
@@ -285,10 +292,12 @@ struct StoreLowering : public OpRewritePattern<vector::StoreOp> {
if (failed(storeLoadPreconditions(rewriter, storeOp, vecTy)))
return failure();
- auto descType =
- xegpu::TensorDescType::get(vecTy.getShape(), vecTy.getElementType(),
- /*array_length=*/1, /*boundary_check=*/true,
- xegpu::MemorySpace::Global);
+ // Boundary check is available only for block instructions.
+ bool boundaryCheck = vecTy.getRank() > 1;
+
+ auto descType = xegpu::TensorDescType::get(
+ vecTy.getShape(), vecTy.getElementType(),
+ /*array_length=*/1, boundaryCheck, xegpu::MemorySpace::Global);
xegpu::CreateNdDescOp ndDesc = createNdDescriptor(
rewriter, loc, descType, storeOp.getBase(), storeOp.getIndices());
diff --git a/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
index e2a506f8ad5abd..7cef17df79dd2a 100644
--- a/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
@@ -12,7 +12,7 @@ func.func @load_1D_vector(%source: memref<8x16x32xf32>, %offset: index) -> vecto
// CHECK: %[[DESC:.+]] = xegpu.create_nd_tdesc
// CHECK-SAME: %[[SRC]][%[[OFFSET]], %[[OFFSET]], %[[OFFSET]]]
// CHECK-SAME: memref<8x16x32xf32> -> !xegpu.tensor_desc<8xf32,
-// CHECK-SAME: boundary_check = true
+// CHECK-SAME: boundary_check = false
// CHECK: %[[VEC:.+]] = xegpu.load_nd %[[DESC]]{{.*}}-> vector<8xf32>
// CHECK: return %[[VEC]]
diff --git a/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
index 3d45407c2486d6..4f069ebc39db37 100644
--- a/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
@@ -14,7 +14,7 @@ func.func @store_1D_vector(%vec: vector<8xf32>,
// CHECK: %[[DESC:.+]] = xegpu.create_nd_tdesc
// CHECK-SAME: %[[SRC]][%[[OFFSET]], %[[OFFSET]], %[[OFFSET]]]
// CHECK-SAME: memref<8x16x32xf32> -> !xegpu.tensor_desc<8xf32,
-// CHECK-SAME: boundary_check = true
+// CHECK-SAME: boundary_check = false
// CHECK: xegpu.store_nd %[[VEC]], %[[DESC]] : vector<8xf32>
// -----
diff --git a/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
index 4841ecbb62e807..497eb86cea8353 100644
--- a/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
@@ -119,6 +119,19 @@ func.func @no_load_out_of_bounds_non_zero_pad(%source: memref<32x64xf32>,
// -----
+func.func @no_load_out_of_bounds_1D_vector(%source: memref<8x16x32xf32>,
+ %offset: index) -> vector<8xf32> {
+ %c0 = arith.constant 0.0 : f32
+ %0 = vector.transfer_read %source[%offset, %offset, %offset], %c0
+ {in_bounds = [false]} : memref<8x16x32xf32>, vector<8xf32>
+ return %0 : vector<8xf32>
+}
+
+// CHECK-LABEL: @no_load_out_of_bounds_1D_vector(
+// CHECK: vector.transfer_read
+
+// -----
+
func.func @no_load_masked(%source : memref<4xf32>,
%offset : index) -> vector<4xf32> {
%c0 = arith.constant 0.0 : f32
diff --git a/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
index 076760fe21dc86..91e3fb3841f6eb 100644
--- a/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
@@ -164,3 +164,16 @@ func.func @no_store_unsupported_map(%vec: vector<8x16xf32>,
// CHECK-LABEL: @no_store_unsupported_map(
// CHECK: vector.transfer_write
+
+// -----
+
+func.func @no_store_out_of_bounds_1D_vector(%vec: vector<8xf32>,
+ %source: memref<8x16x32xf32>, %offset: index) {
+ vector.transfer_write %vec, %source[%offset, %offset, %offset]
+ {in_bounds = [false]}
+ : vector<8xf32>, memref<8x16x32xf32>
+ return
+}
+
+// CHECK-LABEL: @no_store_out_of_bounds_1D_vector(
+// CHECK: vector.transfer_write
|
Member
|
@llvm/pr-subscribers-mlir Author: Adam Siemieniuk (adam-smnk) ChangesConstrains Vector lowering to apply boundary checks only to data transfers operating on block shapes. This further aligns lowering with the current Xe instructions' restrictions. Full diff: https://github.com/llvm/llvm-project/pull/119380.diff 5 Files Affected:
diff --git a/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp b/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
index 215e1b1b874520..1232d8795d4dcf 100644
--- a/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
+++ b/mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
@@ -82,6 +82,10 @@ static LogicalResult transferPreconditions(PatternRewriter &rewriter,
xferOp, "Buffer must be contiguous in the innermost dimension");
unsigned vecRank = vecTy.getRank();
+ if (xferOp.hasOutOfBoundsDim() && vecRank < 2)
+ return rewriter.notifyMatchFailure(
+ xferOp, "Boundary check is available only for block instructions.");
+
AffineMap map = xferOp.getPermutationMap();
if (!map.isProjectedPermutation(/*allowZeroInResults=*/false))
return rewriter.notifyMatchFailure(xferOp, "Unsupported permutation map");
@@ -255,9 +259,12 @@ struct LoadLowering : public OpRewritePattern<vector::LoadOp> {
if (failed(storeLoadPreconditions(rewriter, loadOp, vecTy)))
return failure();
+ // Boundary check is available only for block instructions.
+ bool boundaryCheck = vecTy.getRank() > 1;
+
auto descType = xegpu::TensorDescType::get(
vecTy.getShape(), vecTy.getElementType(), /*array_length=*/1,
- /*boundary_check=*/true, xegpu::MemorySpace::Global);
+ boundaryCheck, xegpu::MemorySpace::Global);
xegpu::CreateNdDescOp ndDesc = createNdDescriptor(
rewriter, loc, descType, loadOp.getBase(), loadOp.getIndices());
@@ -285,10 +292,12 @@ struct StoreLowering : public OpRewritePattern<vector::StoreOp> {
if (failed(storeLoadPreconditions(rewriter, storeOp, vecTy)))
return failure();
- auto descType =
- xegpu::TensorDescType::get(vecTy.getShape(), vecTy.getElementType(),
- /*array_length=*/1, /*boundary_check=*/true,
- xegpu::MemorySpace::Global);
+ // Boundary check is available only for block instructions.
+ bool boundaryCheck = vecTy.getRank() > 1;
+
+ auto descType = xegpu::TensorDescType::get(
+ vecTy.getShape(), vecTy.getElementType(),
+ /*array_length=*/1, boundaryCheck, xegpu::MemorySpace::Global);
xegpu::CreateNdDescOp ndDesc = createNdDescriptor(
rewriter, loc, descType, storeOp.getBase(), storeOp.getIndices());
diff --git a/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
index e2a506f8ad5abd..7cef17df79dd2a 100644
--- a/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/load-to-xegpu.mlir
@@ -12,7 +12,7 @@ func.func @load_1D_vector(%source: memref<8x16x32xf32>, %offset: index) -> vecto
// CHECK: %[[DESC:.+]] = xegpu.create_nd_tdesc
// CHECK-SAME: %[[SRC]][%[[OFFSET]], %[[OFFSET]], %[[OFFSET]]]
// CHECK-SAME: memref<8x16x32xf32> -> !xegpu.tensor_desc<8xf32,
-// CHECK-SAME: boundary_check = true
+// CHECK-SAME: boundary_check = false
// CHECK: %[[VEC:.+]] = xegpu.load_nd %[[DESC]]{{.*}}-> vector<8xf32>
// CHECK: return %[[VEC]]
diff --git a/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
index 3d45407c2486d6..4f069ebc39db37 100644
--- a/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/store-to-xegpu.mlir
@@ -14,7 +14,7 @@ func.func @store_1D_vector(%vec: vector<8xf32>,
// CHECK: %[[DESC:.+]] = xegpu.create_nd_tdesc
// CHECK-SAME: %[[SRC]][%[[OFFSET]], %[[OFFSET]], %[[OFFSET]]]
// CHECK-SAME: memref<8x16x32xf32> -> !xegpu.tensor_desc<8xf32,
-// CHECK-SAME: boundary_check = true
+// CHECK-SAME: boundary_check = false
// CHECK: xegpu.store_nd %[[VEC]], %[[DESC]] : vector<8xf32>
// -----
diff --git a/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
index 4841ecbb62e807..497eb86cea8353 100644
--- a/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
@@ -119,6 +119,19 @@ func.func @no_load_out_of_bounds_non_zero_pad(%source: memref<32x64xf32>,
// -----
+func.func @no_load_out_of_bounds_1D_vector(%source: memref<8x16x32xf32>,
+ %offset: index) -> vector<8xf32> {
+ %c0 = arith.constant 0.0 : f32
+ %0 = vector.transfer_read %source[%offset, %offset, %offset], %c0
+ {in_bounds = [false]} : memref<8x16x32xf32>, vector<8xf32>
+ return %0 : vector<8xf32>
+}
+
+// CHECK-LABEL: @no_load_out_of_bounds_1D_vector(
+// CHECK: vector.transfer_read
+
+// -----
+
func.func @no_load_masked(%source : memref<4xf32>,
%offset : index) -> vector<4xf32> {
%c0 = arith.constant 0.0 : f32
diff --git a/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir b/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
index 076760fe21dc86..91e3fb3841f6eb 100644
--- a/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
+++ b/mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
@@ -164,3 +164,16 @@ func.func @no_store_unsupported_map(%vec: vector<8x16xf32>,
// CHECK-LABEL: @no_store_unsupported_map(
// CHECK: vector.transfer_write
+
+// -----
+
+func.func @no_store_out_of_bounds_1D_vector(%vec: vector<8xf32>,
+ %source: memref<8x16x32xf32>, %offset: index) {
+ vector.transfer_write %vec, %source[%offset, %offset, %offset]
+ {in_bounds = [false]}
+ : vector<8xf32>, memref<8x16x32xf32>
+ return
+}
+
+// CHECK-LABEL: @no_store_out_of_bounds_1D_vector(
+// CHECK: vector.transfer_write
|
Contributor
Author
|
Follow up on #110811 |
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.
Constrains Vector lowering to apply boundary checks only to data transfers operating on block shapes.
This further aligns lowering with the current Xe instructions' restrictions.