[mlir][vector] Generalize folding of ext-contractionOp to other types. #96593
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
Many state of the art models and quantization operations are now directly working on vector.contract on integers. This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines. Signed-off-by: Stanley Winata <[email protected]>
|
@llvm/pr-subscribers-mlir-vector Author: Stanley Winata (raikonenfnu) ChangesMany state of the art models and quantization operations are now directly working on vector.contract on integers. This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines. Full diff: https://github.com/llvm/llvm-project/pull/96593.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
index ea4a02f2f2e77..6dc0e1c1b4bd8 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
@@ -1542,6 +1542,7 @@ struct CanonicalizeContractMatmulToMMT final
/// Cores, i.e, `mma.sync.*.f32.f16.f16.f32` and `mma.sync.*.f32.bf16.bf16.f32`.
/// This pattern folds the arithmetic extensions into the vector contraction and
/// enables the usage of native mixed precision Tensor Core instructions.
+template <typename ExtOp>
struct FoldArithExtIntoContractionOp
: public OpRewritePattern<vector::ContractionOp> {
using OpRewritePattern::OpRewritePattern;
@@ -1549,8 +1550,8 @@ struct FoldArithExtIntoContractionOp
LogicalResult matchAndRewrite(vector::ContractionOp contractOp,
PatternRewriter &rewriter) const override {
- auto lhsDefOp = contractOp.getLhs().getDefiningOp<arith::ExtFOp>();
- auto rhsDefOp = contractOp.getRhs().getDefiningOp<arith::ExtFOp>();
+ auto lhsDefOp = contractOp.getLhs().getDefiningOp<ExtOp>();
+ auto rhsDefOp = contractOp.getRhs().getDefiningOp<ExtOp>();
if (!lhsDefOp || !rhsDefOp) {
return rewriter.notifyMatchFailure(contractOp,
@@ -1804,7 +1805,9 @@ struct BreakDownVectorReduction final : OpRewritePattern<vector::ReductionOp> {
void mlir::vector::populateFoldArithExtensionPatterns(
RewritePatternSet &patterns) {
- patterns.add<FoldArithExtIntoContractionOp>(patterns.getContext());
+ patterns.add<FoldArithExtIntoContractionOp<arith::ExtFOp>,
+ FoldArithExtIntoContractionOp<arith::ExtSIOp>>(
+ patterns.getContext());
}
void mlir::vector::populateVectorMaskMaterializationPatterns(
diff --git a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
index 31ae126906f21..6dbde7afbdd33 100644
--- a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
+++ b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
@@ -48,3 +48,25 @@ func.func @fold_arith_extf_into_contract_scalable(
%lhs_f32, %rhs_f32, %arg2 : vector<[64]x64xf32>, vector<64x64xf32> into vector<[64]x64xf32>
return %result : vector<[64]x64xf32>
}
+
+// -----
+
+// CHECK-LABEL: func.func @fold_arith_extsi_into_contract
+// CHECK-SAME: (%[[ARG0:.*]]: vector<64x64xi8>, %[[ARG1:.*]]: vector<64x64xi8>, %[[ARG2:.*]]: vector<64x64xi32>)
+// CHECK-NEXT: %[[R:.+]] = vector.contract
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "reduction"], kind = #vector.kind<add>}
+// CHECK-SAME: %[[ARG0]], %[[ARG1]], %[[ARG2]] : vector<64x64xi8>, vector<64x64xi8> into vector<64x64xi32>
+// CHECK-NEXT: return %[[R]] : vector<64x64xi32>
+func.func @fold_arith_extsi_into_contract(
+ %arg0: vector<64x64xi8>,
+ %arg1: vector<64x64xi8>,
+ %arg2: vector<64x64xi32>) -> vector<64x64xi32> {
+ %lhs_i32 = arith.extsi %arg0 : vector<64x64xi8> to vector<64x64xi32>
+ %rhs_i32 = arith.extsi %arg1 : vector<64x64xi8> to vector<64x64xi32>
+ %result = vector.contract {
+ indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d2)>, affine_map<(d0, d1, d2) -> (d2, d1)>, affine_map<(d0, d1, d2) -> (d0, d1)>],
+ iterator_types = ["parallel", "parallel", "reduction"],
+ kind = #vector.kind<add>}
+ %lhs_i32, %rhs_i32, %arg2 : vector<64x64xi32>, vector<64x64xi32> into vector<64x64xi32>
+ return %result : vector<64x64xi32>
+}
|
|
@llvm/pr-subscribers-mlir Author: Stanley Winata (raikonenfnu) ChangesMany state of the art models and quantization operations are now directly working on vector.contract on integers. This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines. Full diff: https://github.com/llvm/llvm-project/pull/96593.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
index ea4a02f2f2e77..6dc0e1c1b4bd8 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
@@ -1542,6 +1542,7 @@ struct CanonicalizeContractMatmulToMMT final
/// Cores, i.e, `mma.sync.*.f32.f16.f16.f32` and `mma.sync.*.f32.bf16.bf16.f32`.
/// This pattern folds the arithmetic extensions into the vector contraction and
/// enables the usage of native mixed precision Tensor Core instructions.
+template <typename ExtOp>
struct FoldArithExtIntoContractionOp
: public OpRewritePattern<vector::ContractionOp> {
using OpRewritePattern::OpRewritePattern;
@@ -1549,8 +1550,8 @@ struct FoldArithExtIntoContractionOp
LogicalResult matchAndRewrite(vector::ContractionOp contractOp,
PatternRewriter &rewriter) const override {
- auto lhsDefOp = contractOp.getLhs().getDefiningOp<arith::ExtFOp>();
- auto rhsDefOp = contractOp.getRhs().getDefiningOp<arith::ExtFOp>();
+ auto lhsDefOp = contractOp.getLhs().getDefiningOp<ExtOp>();
+ auto rhsDefOp = contractOp.getRhs().getDefiningOp<ExtOp>();
if (!lhsDefOp || !rhsDefOp) {
return rewriter.notifyMatchFailure(contractOp,
@@ -1804,7 +1805,9 @@ struct BreakDownVectorReduction final : OpRewritePattern<vector::ReductionOp> {
void mlir::vector::populateFoldArithExtensionPatterns(
RewritePatternSet &patterns) {
- patterns.add<FoldArithExtIntoContractionOp>(patterns.getContext());
+ patterns.add<FoldArithExtIntoContractionOp<arith::ExtFOp>,
+ FoldArithExtIntoContractionOp<arith::ExtSIOp>>(
+ patterns.getContext());
}
void mlir::vector::populateVectorMaskMaterializationPatterns(
diff --git a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
index 31ae126906f21..6dbde7afbdd33 100644
--- a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
+++ b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
@@ -48,3 +48,25 @@ func.func @fold_arith_extf_into_contract_scalable(
%lhs_f32, %rhs_f32, %arg2 : vector<[64]x64xf32>, vector<64x64xf32> into vector<[64]x64xf32>
return %result : vector<[64]x64xf32>
}
+
+// -----
+
+// CHECK-LABEL: func.func @fold_arith_extsi_into_contract
+// CHECK-SAME: (%[[ARG0:.*]]: vector<64x64xi8>, %[[ARG1:.*]]: vector<64x64xi8>, %[[ARG2:.*]]: vector<64x64xi32>)
+// CHECK-NEXT: %[[R:.+]] = vector.contract
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "reduction"], kind = #vector.kind<add>}
+// CHECK-SAME: %[[ARG0]], %[[ARG1]], %[[ARG2]] : vector<64x64xi8>, vector<64x64xi8> into vector<64x64xi32>
+// CHECK-NEXT: return %[[R]] : vector<64x64xi32>
+func.func @fold_arith_extsi_into_contract(
+ %arg0: vector<64x64xi8>,
+ %arg1: vector<64x64xi8>,
+ %arg2: vector<64x64xi32>) -> vector<64x64xi32> {
+ %lhs_i32 = arith.extsi %arg0 : vector<64x64xi8> to vector<64x64xi32>
+ %rhs_i32 = arith.extsi %arg1 : vector<64x64xi8> to vector<64x64xi32>
+ %result = vector.contract {
+ indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d2)>, affine_map<(d0, d1, d2) -> (d2, d1)>, affine_map<(d0, d1, d2) -> (d0, d1)>],
+ iterator_types = ["parallel", "parallel", "reduction"],
+ kind = #vector.kind<add>}
+ %lhs_i32, %rhs_i32, %arg2 : vector<64x64xi32>, vector<64x64xi32> into vector<64x64xi32>
+ return %result : vector<64x64xi32>
+}
|
MaheshRavishankar
approved these changes
Jun 25, 2024
kuhar
pushed a commit
to iree-org/llvm-project
that referenced
this pull request
Jun 27, 2024
llvm#96593) Many state of the art models and quantization operations are now directly working on vector.contract on integers. This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines. Signed-off-by: Stanley Winata <[email protected]>
AlexisPerry
pushed a commit
to llvm-project-tlp/llvm-project
that referenced
this pull request
Jul 9, 2024
llvm#96593) Many state of the art models and quantization operations are now directly working on vector.contract on integers. This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines. 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
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.
Many state of the art models and quantization operations are now directly working on vector.contract on integers.
This commit enables generalizes ext-contraction folding S.T we can emit more performant vector.contracts on codegen pipelines.