[Mlir-commits] [mlir] [mlir][vector] Make the in_bounds attribute mandatory (PR #97049)
Andrzej Warzyński
llvmlistbot at llvm.org
Fri Jun 28 05:48:35 PDT 2024
https://github.com/banach-space created https://github.com/llvm/llvm-project/pull/97049
WIP
>From 9e1cf900649d4a99671f2fe39393b8212acd2cb4 Mon Sep 17 00:00:00 2001
From: Andrzej Warzynski <andrzej.warzynski at arm.com>
Date: Wed, 19 Jun 2024 07:57:59 +0100
Subject: [PATCH] [mlir][vector] Make the in_bounds attribute mandatory
WIP
---
.../mlir/Dialect/Vector/IR/VectorOps.td | 4 +-
.../mlir/Interfaces/VectorInterfaces.td | 4 +-
mlir/lib/Dialect/Vector/IR/VectorOps.cpp | 7 +-
.../Vector/Transforms/LowerVectorMask.cpp | 4 +-
.../Vector/Transforms/LowerVectorTransfer.cpp | 10 +--
.../Conversion/VectorToSCF/vector-to-scf.mlir | 2 +-
.../Linalg/vectorization-with-patterns.mlir | 6 +-
mlir/test/Dialect/Vector/ops.mlir | 4 +-
.../loop-invariant-subset-hoisting.mlir | 78 +++++++++----------
9 files changed, 56 insertions(+), 63 deletions(-)
diff --git a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
index 56d866ac5b40c..4e4d53582ae30 100644
--- a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
+++ b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
@@ -1321,7 +1321,7 @@ def Vector_TransferReadOp :
AffineMapAttr:$permutation_map,
AnyType:$padding,
Optional<VectorOf<[I1]>>:$mask,
- OptionalAttr<BoolArrayAttr>:$in_bounds)>,
+ BoolArrayAttr:$in_bounds)>,
Results<(outs AnyVectorOfAnyRank:$vector)> {
let summary = "Reads a supervector from memory into an SSA vector value.";
@@ -1565,7 +1565,7 @@ def Vector_TransferWriteOp :
Variadic<Index>:$indices,
AffineMapAttr:$permutation_map,
Optional<VectorOf<[I1]>>:$mask,
- OptionalAttr<BoolArrayAttr>:$in_bounds)>,
+ BoolArrayAttr:$in_bounds)>,
Results<(outs Optional<AnyRankedTensor>:$result)> {
let summary = "The vector.transfer_write op writes a supervector to memory.";
diff --git a/mlir/include/mlir/Interfaces/VectorInterfaces.td b/mlir/include/mlir/Interfaces/VectorInterfaces.td
index 781d6d3e3f813..f6682f2eabe1e 100644
--- a/mlir/include/mlir/Interfaces/VectorInterfaces.td
+++ b/mlir/include/mlir/Interfaces/VectorInterfaces.td
@@ -98,7 +98,7 @@ def VectorTransferOpInterface : OpInterface<"VectorTransferOpInterface"> {
dimension whether it is in-bounds or not. (Broadcast dimensions are
always in-bounds).
}],
- /*retTy=*/"::std::optional<::mlir::ArrayAttr>",
+ /*retTy=*/"::mlir::ArrayAttr",
/*methodName=*/"getInBounds",
/*args=*/(ins)
>,
@@ -242,7 +242,7 @@ def VectorTransferOpInterface : OpInterface<"VectorTransferOpInterface"> {
return true;
if (!$_op.getInBounds())
return false;
- auto inBounds = ::llvm::cast<::mlir::ArrayAttr>(*$_op.getInBounds());
+ auto inBounds = $_op.getInBounds();
return ::llvm::cast<::mlir::BoolAttr>(inBounds[dim]).getValue();
}
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index 2bf4f16f96e6a..90a83ebe5c8aa 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -3859,9 +3859,6 @@ static void printTransferAttrs(OpAsmPrinter &p, VectorTransferOpInterface op) {
elidedAttrs.push_back(TransferReadOp::getOperandSegmentSizeAttr());
if (op.getPermutationMap().isMinorIdentity())
elidedAttrs.push_back(op.getPermutationMapAttrName());
- // Elide in_bounds attribute if all dims are out-of-bounds.
- if (llvm::none_of(op.getInBoundsValues(), [](bool b) { return b; }))
- elidedAttrs.push_back(op.getInBoundsAttrName());
p.printOptionalAttrDict(op->getAttrs(), elidedAttrs);
}
@@ -3970,7 +3967,7 @@ LogicalResult TransferReadOp::verify() {
if (failed(verifyTransferOp(cast<VectorTransferOpInterface>(getOperation()),
shapedType, vectorType, maskType,
inferredMaskType, permutationMap,
- getInBounds() ? *getInBounds() : ArrayAttr())))
+ getInBounds())))
return failure();
if (auto sourceVectorElementType =
@@ -4356,7 +4353,7 @@ LogicalResult TransferWriteOp::verify() {
if (failed(verifyTransferOp(cast<VectorTransferOpInterface>(getOperation()),
shapedType, vectorType, maskType,
inferredMaskType, permutationMap,
- getInBounds() ? *getInBounds() : ArrayAttr())))
+ getInBounds())))
return failure();
return verifyPermutationMap(permutationMap,
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorMask.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorMask.cpp
index f53bb5157eb37..dfeb7bc53adad 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorMask.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorMask.cpp
@@ -224,7 +224,7 @@ struct MaskedTransferReadOpPattern
rewriter.replaceOpWithNewOp<TransferReadOp>(
maskingOp.getOperation(), readOp.getVectorType(), readOp.getSource(),
readOp.getIndices(), readOp.getPermutationMap(), readOp.getPadding(),
- maskingOp.getMask(), readOp.getInBounds().value_or(ArrayAttr()));
+ maskingOp.getMask(), readOp.getInBounds());
return success();
}
};
@@ -246,7 +246,7 @@ struct MaskedTransferWriteOpPattern
rewriter.replaceOpWithNewOp<TransferWriteOp>(
maskingOp.getOperation(), resultType, writeOp.getVector(),
writeOp.getSource(), writeOp.getIndices(), writeOp.getPermutationMap(),
- maskingOp.getMask(), writeOp.getInBounds().value_or(ArrayAttr()));
+ maskingOp.getMask(), writeOp.getInBounds());
return success();
}
};
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
index c31c51489ecc9..a8a6fef284907 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
@@ -132,10 +132,8 @@ struct TransferReadPermutationLowering
}
// Transpose in_bounds attribute.
- ArrayAttr newInBoundsAttr =
- op.getInBounds() ? inverseTransposeInBoundsAttr(
- rewriter, op.getInBounds().value(), permutation)
- : ArrayAttr();
+ ArrayAttr newInBoundsAttr = inverseTransposeInBoundsAttr(
+ rewriter, op.getInBounds(), permutation);
// Generate new transfer_read operation.
VectorType newReadType = VectorType::get(
@@ -208,9 +206,7 @@ struct TransferWritePermutationLowering
// Transpose in_bounds attribute.
ArrayAttr newInBoundsAttr =
- op.getInBounds() ? inverseTransposeInBoundsAttr(
- rewriter, op.getInBounds().value(), permutation)
- : ArrayAttr();
+ inverseTransposeInBoundsAttr(rewriter, op.getInBounds(), permutation);
// Generate new transfer_write operation.
Value newVec = rewriter.create<vector::TransposeOp>(
diff --git a/mlir/test/Conversion/VectorToSCF/vector-to-scf.mlir b/mlir/test/Conversion/VectorToSCF/vector-to-scf.mlir
index e1babdd2f1f63..ea57e2afbaa2b 100644
--- a/mlir/test/Conversion/VectorToSCF/vector-to-scf.mlir
+++ b/mlir/test/Conversion/VectorToSCF/vector-to-scf.mlir
@@ -174,7 +174,7 @@ func.func @materialize_write(%M: index, %N: index, %O: index, %P: index) {
// CHECK: scf.for %[[I6:.*]] = %[[C0]] to %[[C1]] step %[[C1]] {
// CHECK: %[[S0:.*]] = affine.apply #[[$ADD]](%[[I2]], %[[I6]])
// CHECK: %[[VEC:.*]] = memref.load %[[VECTOR_VIEW3]][%[[I4]], %[[I5]], %[[I6]]] : memref<3x4x1xvector<5xf32>>
- // CHECK: vector.transfer_write %[[VEC]], %{{.*}}[%[[S3]], %[[S1]], %[[S0]], %[[I3]]] : vector<5xf32>, memref<?x?x?x?xf32>
+ // CHECK: vector.transfer_write %[[VEC]], %{{.*}}[%[[S3]], %[[S1]], %[[S0]], %[[I3]]] {in_bounds = [false]} : vector<5xf32>, memref<?x?x?x?xf32>
// CHECK: }
// CHECK: }
// CHECK: }
diff --git a/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir b/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
index d7ff1ded9d933..7176dff9bc857 100644
--- a/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
@@ -950,7 +950,7 @@ module attributes {transform.with_named_sequence} {
// CHECK-NOT: tensor.pad
// CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[C5:.*]] = arith.constant 5.0
-// CHECK: %[[RESULT:.*]] = vector.transfer_read %[[ARG0]][%[[C0]], %[[C0]]], %[[C5]] : tensor<5x6xf32>, vector<7x9xf32>
+// CHECK: %[[RESULT:.*]] = vector.transfer_read %[[ARG0]][%[[C0]], %[[C0]]], %[[C5]] {in_bounds = [false, false]} : tensor<5x6xf32>, vector<7x9xf32>
// CHECK: return %[[RESULT]]
func.func @pad_and_transfer_read(%arg0: tensor<5x6xf32>) -> vector<7x9xf32> {
%c0 = arith.constant 0 : index
@@ -984,7 +984,7 @@ func.func private @make_vector() -> vector<7x9xf32>
// CHECK-NOT: tensor.pad
// CHECK: %[[C0:.*]] = arith.constant 0 : index
// CHECK: %[[VEC0:.*]] = call @make_vector() : () -> vector<7x9xf32>
-// CHECK: %[[RESULT:.*]] = vector.transfer_write %[[VEC0]], %[[ARG0]][%[[C0]], %[[C0]]] : vector<7x9xf32>, tensor<5x6xf32>
+// CHECK: %[[RESULT:.*]] = vector.transfer_write %[[VEC0]], %[[ARG0]][%[[C0]], %[[C0]]] {in_bounds = [false, false]} : vector<7x9xf32>, tensor<5x6xf32>
// CHECK: return %[[RESULT]]
func.func @pad_and_transfer_write_static(
%arg0: tensor<5x6xf32>) -> tensor<5x6xf32> {
@@ -1021,7 +1021,7 @@ func.func private @make_vector() -> vector<7x9xf32>
// CHECK: %[[C0:.*]] = arith.constant 0 : index
// CHECK: %[[SUB:.*]] = tensor.extract_slice %[[ARG0]][0, 0] [%[[SIZE]], 6] [1, 1] : tensor<?x?xf32> to tensor<?x6xf32>
// CHECK: %[[VEC0:.*]] = call @make_vector() : () -> vector<7x9xf32>
-// CHECK: %[[RESULT:.*]] = vector.transfer_write %[[VEC0]], %[[SUB]][%[[C0]], %[[C0]]] : vector<7x9xf32>, tensor<?x6xf32>
+// CHECK: %[[RESULT:.*]] = vector.transfer_write %[[VEC0]], %[[SUB]][%[[C0]], %[[C0]]] {in_bounds = [false, false]} : vector<7x9xf32>, tensor<?x6xf32>
// CHECK: return %[[RESULT]]
func.func @pad_and_transfer_write_dynamic_static(
%arg0: tensor<?x?xf32>, %size: index, %padding: index) -> tensor<?x6xf32> {
diff --git a/mlir/test/Dialect/Vector/ops.mlir b/mlir/test/Dialect/Vector/ops.mlir
index c868c881d079a..afe62a2427fb0 100644
--- a/mlir/test/Dialect/Vector/ops.mlir
+++ b/mlir/test/Dialect/Vector/ops.mlir
@@ -78,7 +78,7 @@ func.func @vector_transfer_ops(%arg0: memref<?x?xf32>,
vector.transfer_write %1, %arg0[%c3, %c3] {permutation_map = affine_map<(d0, d1)->(d1, d0)>} : vector<3x7xf32>, memref<?x?xf32>
// CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<1x1x4x3xf32>, memref<?x?xvector<4x3xf32>>
vector.transfer_write %4, %arg1[%c3, %c3] {permutation_map = affine_map<(d0, d1)->(d0, d1)>} : vector<1x1x4x3xf32>, memref<?x?xvector<4x3xf32>>
- // CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<1x1x4x3xf32>, memref<?x?xvector<4x3xf32>>
+ // CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] {in_bounds = [false, false]} : vector<1x1x4x3xf32>, memref<?x?xvector<4x3xf32>>
vector.transfer_write %5, %arg1[%c3, %c3] {in_bounds = [false, false]} : vector<1x1x4x3xf32>, memref<?x?xvector<4x3xf32>>
// CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<5x24xi8>, memref<?x?xvector<4x3xi32>>
vector.transfer_write %6, %arg2[%c3, %c3] : vector<5x24xi8>, memref<?x?xvector<4x3xi32>>
@@ -135,7 +135,7 @@ func.func @vector_transfer_ops_tensor(%arg0: tensor<?x?xf32>,
%9 = vector.transfer_write %1, %arg0[%c3, %c3] {permutation_map = affine_map<(d0, d1)->(d1, d0)>} : vector<3x7xf32>, tensor<?x?xf32>
// CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<1x1x4x3xf32>, tensor<?x?xvector<4x3xf32>>
%10 = vector.transfer_write %4, %arg1[%c3, %c3] {permutation_map = affine_map<(d0, d1)->(d0, d1)>} : vector<1x1x4x3xf32>, tensor<?x?xvector<4x3xf32>>
- // CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<1x1x4x3xf32>, tensor<?x?xvector<4x3xf32>>
+ // CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] {in_bounds = [false, false]} : vector<1x1x4x3xf32>, tensor<?x?xvector<4x3xf32>>
%11 = vector.transfer_write %5, %arg1[%c3, %c3] {in_bounds = [false, false]} : vector<1x1x4x3xf32>, tensor<?x?xvector<4x3xf32>>
// CHECK: vector.transfer_write %{{.*}}, %{{.*}}[%[[C3]], %[[C3]]] : vector<5x24xi8>, tensor<?x?xvector<4x3xi32>>
%12 = vector.transfer_write %6, %arg2[%c3, %c3] : vector<5x24xi8>, tensor<?x?xvector<4x3xi32>>
diff --git a/mlir/test/Transforms/loop-invariant-subset-hoisting.mlir b/mlir/test/Transforms/loop-invariant-subset-hoisting.mlir
index 3a78287a0dcad..54e2d92f9a952 100644
--- a/mlir/test/Transforms/loop-invariant-subset-hoisting.mlir
+++ b/mlir/test/Transforms/loop-invariant-subset-hoisting.mlir
@@ -326,12 +326,12 @@ func.func @hoist_vector_transfer_pairs_tensor(
%arg9 = %arg3, %arg10 = %arg4, %arg11 = %arg5)
-> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>,
tensor<?x?xf32>, tensor<?x?xf32>) {
- %r0 = vector.transfer_read %arg7[%c0, %c0], %cst: tensor<?x?xf32>, vector<1xf32>
- %r1 = vector.transfer_read %arg6[%i, %i], %cst: tensor<?x?xf32>, vector<2xf32>
- %r3 = vector.transfer_read %arg9[%c0, %c0], %cst: tensor<?x?xf32>, vector<4xf32>
+ %r0 = vector.transfer_read %arg7[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<1xf32>
+ %r1 = vector.transfer_read %arg6[%i, %i], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
+ %r3 = vector.transfer_read %arg9[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<4xf32>
"test.some_crippling_use"(%arg10) : (tensor<?x?xf32>) -> ()
- %r4 = vector.transfer_read %arg10[%c0, %c0], %cst: tensor<?x?xf32>, vector<5xf32>
- %r5 = vector.transfer_read %arg11[%c0, %c0], %cst: tensor<?x?xf32>, vector<6xf32>
+ %r4 = vector.transfer_read %arg10[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<5xf32>
+ %r5 = vector.transfer_read %arg11[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<6xf32>
"test.some_crippling_use"(%arg11) : (tensor<?x?xf32>) -> ()
%u0 = "test.some_use"(%r0) : (vector<1xf32>) -> vector<1xf32>
%u1 = "test.some_use"(%r1) : (vector<2xf32>) -> vector<2xf32>
@@ -339,12 +339,12 @@ func.func @hoist_vector_transfer_pairs_tensor(
%u3 = "test.some_use"(%r3) : (vector<4xf32>) -> vector<4xf32>
%u4 = "test.some_use"(%r4) : (vector<5xf32>) -> vector<5xf32>
%u5 = "test.some_use"(%r5) : (vector<6xf32>) -> vector<6xf32>
- %w1 = vector.transfer_write %u0, %arg7[%c0, %c0] : vector<1xf32>, tensor<?x?xf32>
- %w0 = vector.transfer_write %u1, %arg6[%i, %i] : vector<2xf32>, tensor<?x?xf32>
- %w2 = vector.transfer_write %u2, %arg8[%c0, %c0] : vector<3xf32>, tensor<?x?xf32>
- %w3 = vector.transfer_write %u3, %arg9[%c0, %c0] : vector<4xf32>, tensor<?x?xf32>
- %w4 = vector.transfer_write %u4, %arg10[%c0, %c0] : vector<5xf32>, tensor<?x?xf32>
- %w5 = vector.transfer_write %u5, %arg11[%c0, %c0] : vector<6xf32>, tensor<?x?xf32>
+ %w1 = vector.transfer_write %u0, %arg7[%c0, %c0] {in_bounds=[true]} : vector<1xf32>, tensor<?x?xf32>
+ %w0 = vector.transfer_write %u1, %arg6[%i, %i] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
+ %w2 = vector.transfer_write %u2, %arg8[%c0, %c0] {in_bounds=[true]} : vector<3xf32>, tensor<?x?xf32>
+ %w3 = vector.transfer_write %u3, %arg9[%c0, %c0] {in_bounds=[true]} : vector<4xf32>, tensor<?x?xf32>
+ %w4 = vector.transfer_write %u4, %arg10[%c0, %c0] {in_bounds=[true]} : vector<5xf32>, tensor<?x?xf32>
+ %w5 = vector.transfer_write %u5, %arg11[%c0, %c0] {in_bounds=[true]} : vector<6xf32>, tensor<?x?xf32>
"test.some_crippling_use"(%w3) : (tensor<?x?xf32>) -> ()
scf.yield %w0, %w1, %w2, %w3, %w4, %w5 :
tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>,
@@ -415,14 +415,14 @@ func.func @hoist_vector_transfer_pairs_disjoint_tensor(
iter_args(%arg4 = %arg0, %arg5 = %arg1, %arg6 = %arg2,
%arg7 = %arg3)
-> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>) {
- %r00 = vector.transfer_read %arg5[%c0, %c0], %cst: tensor<?x?xf32>, vector<2xf32>
- %r01 = vector.transfer_read %arg5[%c0, %c1], %cst: tensor<?x?xf32>, vector<2xf32>
- %r20 = vector.transfer_read %arg6[%c0, %c0], %cst: tensor<?x?xf32>, vector<3xf32>
- %r21 = vector.transfer_read %arg6[%c0, %c3], %cst: tensor<?x?xf32>, vector<3xf32>
- %r30 = vector.transfer_read %arg7[%c0, %random_index], %cst: tensor<?x?xf32>, vector<4xf32>
- %r31 = vector.transfer_read %arg7[%c1, %random_index], %cst: tensor<?x?xf32>, vector<4xf32>
- %r10 = vector.transfer_read %arg4[%i, %i], %cst: tensor<?x?xf32>, vector<2xf32>
- %r11 = vector.transfer_read %arg4[%random_index, %random_index], %cst: tensor<?x?xf32>, vector<2xf32>
+ %r00 = vector.transfer_read %arg5[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
+ %r01 = vector.transfer_read %arg5[%c0, %c1], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
+ %r20 = vector.transfer_read %arg6[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<3xf32>
+ %r21 = vector.transfer_read %arg6[%c0, %c3], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<3xf32>
+ %r30 = vector.transfer_read %arg7[%c0, %random_index], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<4xf32>
+ %r31 = vector.transfer_read %arg7[%c1, %random_index], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<4xf32>
+ %r10 = vector.transfer_read %arg4[%i, %i], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
+ %r11 = vector.transfer_read %arg4[%random_index, %random_index], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
%u00 = "test.some_use"(%r00) : (vector<2xf32>) -> vector<2xf32>
%u01 = "test.some_use"(%r01) : (vector<2xf32>) -> vector<2xf32>
%u20 = "test.some_use"(%r20) : (vector<3xf32>) -> vector<3xf32>
@@ -431,14 +431,14 @@ func.func @hoist_vector_transfer_pairs_disjoint_tensor(
%u31 = "test.some_use"(%r31) : (vector<4xf32>) -> vector<4xf32>
%u10 = "test.some_use"(%r10) : (vector<2xf32>) -> vector<2xf32>
%u11 = "test.some_use"(%r11) : (vector<2xf32>) -> vector<2xf32>
- %w10 = vector.transfer_write %u00, %arg5[%c0, %c0] : vector<2xf32>, tensor<?x?xf32>
- %w11 = vector.transfer_write %u01, %w10[%c0, %c1] : vector<2xf32>, tensor<?x?xf32>
- %w20 = vector.transfer_write %u20, %arg6[%c0, %c0] : vector<3xf32>, tensor<?x?xf32>
- %w21 = vector.transfer_write %u21, %w20[%c0, %c3] : vector<3xf32>, tensor<?x?xf32>
- %w30 = vector.transfer_write %u30, %arg7[%c0, %random_index] : vector<4xf32>, tensor<?x?xf32>
- %w31 = vector.transfer_write %u31, %w30[%c1, %random_index] : vector<4xf32>, tensor<?x?xf32>
- %w00 = vector.transfer_write %u10, %arg4[%i, %i] : vector<2xf32>, tensor<?x?xf32>
- %w01 = vector.transfer_write %u11, %w00[%random_index, %random_index] : vector<2xf32>, tensor<?x?xf32>
+ %w10 = vector.transfer_write %u00, %arg5[%c0, %c0] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
+ %w11 = vector.transfer_write %u01, %w10[%c0, %c1] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
+ %w20 = vector.transfer_write %u20, %arg6[%c0, %c0] {in_bounds=[true]} : vector<3xf32>, tensor<?x?xf32>
+ %w21 = vector.transfer_write %u21, %w20[%c0, %c3] {in_bounds=[true]} : vector<3xf32>, tensor<?x?xf32>
+ %w30 = vector.transfer_write %u30, %arg7[%c0, %random_index] {in_bounds=[true]} : vector<4xf32>, tensor<?x?xf32>
+ %w31 = vector.transfer_write %u31, %w30[%c1, %random_index] {in_bounds=[true]} : vector<4xf32>, tensor<?x?xf32>
+ %w00 = vector.transfer_write %u10, %arg4[%i, %i] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
+ %w01 = vector.transfer_write %u11, %w00[%random_index, %random_index] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
scf.yield %w01, %w11, %w21, %w31 : tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>
}
scf.yield %1#0, %1#1, %1#2, %1#3 : tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>
@@ -492,19 +492,19 @@ func.func @hoist_vector_transfer_pairs_tensor_and_slices(
-> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>) {
// Hoists.
%st0 = tensor.extract_slice %arg6[%i, %i][%step, %step][1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
- %r0 = vector.transfer_read %st0[%c0, %c0], %cst: tensor<?x?xf32>, vector<1xf32>
+ %r0 = vector.transfer_read %st0[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<1xf32>
// CHECK: %[[ST1:.*]] = tensor.extract_slice %[[TENSOR1_ARG_L2]][%[[J]],{{.*}}: tensor<?x?xf32> to tensor<?x?xf32>
// CHECK: %[[V1:.*]] = vector.transfer_read %[[ST1]]{{.*}} : tensor<?x?xf32>, vector<2xf32>
// Does not hoist (slice depends on %j)
%st1 = tensor.extract_slice %arg7[%j, %c0][%step, %step][1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
- %r1 = vector.transfer_read %st1[%c0, %c0], %cst: tensor<?x?xf32>, vector<2xf32>
+ %r1 = vector.transfer_read %st1[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
// CHECK: %[[ST2:.*]] = tensor.extract_slice %[[TENSOR2_ARG_L2]][%[[I]],{{.*}}: tensor<?x?xf32> to tensor<?x?xf32>
// CHECK: %[[V2:.*]] = vector.transfer_read %[[ST2]]{{.*}} : tensor<?x?xf32>, vector<3xf32>
// Does not hoist, 2 slice %arg8.
%st2 = tensor.extract_slice %arg8[%i, %c0][%step, %step][1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
- %r2 = vector.transfer_read %st2[%c0, %c0], %cst: tensor<?x?xf32>, vector<3xf32>
+ %r2 = vector.transfer_read %st2[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<3xf32>
// CHECK: %[[U0:.*]] = "test.some_use"(%[[V0_ARG_L2]]) : (vector<1xf32>) -> vector<1xf32>
// CHECK: %[[U1:.*]] = "test.some_use"(%[[V1]]) : (vector<2xf32>) -> vector<2xf32>
@@ -514,15 +514,15 @@ func.func @hoist_vector_transfer_pairs_tensor_and_slices(
%u2 = "test.some_use"(%r2) : (vector<3xf32>) -> vector<3xf32>
// Hoists
- %w0 = vector.transfer_write %u0, %st0[%c0, %c0] : vector<1xf32>, tensor<?x?xf32>
+ %w0 = vector.transfer_write %u0, %st0[%c0, %c0] {in_bounds=[true]} : vector<1xf32>, tensor<?x?xf32>
// CHECK-DAG: %[[STI1:.*]] = vector.transfer_write %[[U1]], %{{.*}} : vector<2xf32>, tensor<?x?xf32>
// Does not hoist (associated slice depends on %j).
- %w1 = vector.transfer_write %u1, %st1[%i, %i] : vector<2xf32>, tensor<?x?xf32>
+ %w1 = vector.transfer_write %u1, %st1[%i, %i] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
// CHECK-DAG: %[[STI2:.*]] = vector.transfer_write %[[U2]], %{{.*}} : vector<3xf32>, tensor<?x?xf32>
// Does not hoist, 2 slice / insert_slice for %arg8.
- %w2 = vector.transfer_write %u2, %st2[%c0, %c0] : vector<3xf32>, tensor<?x?xf32>
+ %w2 = vector.transfer_write %u2, %st2[%c0, %c0] {in_bounds=[true]} : vector<3xf32>, tensor<?x?xf32>
// Hoists.
%sti0 = tensor.insert_slice %w0 into %arg6[%i, %i][%step, %step][1, 1] : tensor<?x?xf32> into tensor<?x?xf32>
@@ -570,8 +570,8 @@ func.func @hoist_vector_transfer_pairs_tensor_and_slices(
// CHECK: %[[R5:.*]] = "test.some_use"(%[[R3]]) : (vector<2xf32>) -> vector<2xf32>
// CHECK: scf.yield %[[TL]], %[[R4]], %[[R5]] : tensor<?x?xf32>, vector<2xf32>, vector<2xf32>
// CHECK: }
-// CHECK: %[[W0:.*]] = vector.transfer_write %[[F]]#2, %[[F]]#0[%[[C0]], %[[C3]]] : vector<2xf32>, tensor<?x?xf32>
-// CHECK: %[[W1:.*]] = vector.transfer_write %[[F]]#1, %[[W0]][%[[C0]], %[[C0]]] : vector<2xf32>, tensor<?x?xf32>
+// CHECK: %[[W0:.*]] = vector.transfer_write %[[F]]#2, %[[F]]#0[%[[C0]], %[[C3]]] {{.*}} : vector<2xf32>, tensor<?x?xf32>
+// CHECK: %[[W1:.*]] = vector.transfer_write %[[F]]#1, %[[W0]][%[[C0]], %[[C0]]] {{.*}} : vector<2xf32>, tensor<?x?xf32>
// CHECK: return %[[W1]] : tensor<?x?xf32>
func.func @hoist_vector_transfer_write_pairs_disjoint_tensor(
%tensor: tensor<?x?xf32>,
@@ -583,14 +583,14 @@ func.func @hoist_vector_transfer_write_pairs_disjoint_tensor(
%cst = arith.constant 0.0 : f32
%1 = scf.for %j = %lb to %ub step %step iter_args(%arg5 = %tensor)
-> (tensor<?x?xf32>) {
- %r00 = vector.transfer_read %arg5[%c0, %c0], %cst: tensor<?x?xf32>, vector<2xf32>
+ %r00 = vector.transfer_read %arg5[%c0, %c0], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
%u00 = "test.some_use"(%r00) : (vector<2xf32>) -> vector<2xf32>
- %w10 = vector.transfer_write %u00, %arg5[%c0, %c0] : vector<2xf32>, tensor<?x?xf32>
+ %w10 = vector.transfer_write %u00, %arg5[%c0, %c0] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
// Hoist by properly bypassing the disjoint write %w10.
- %r01 = vector.transfer_read %w10[%c0, %c3], %cst: tensor<?x?xf32>, vector<2xf32>
+ %r01 = vector.transfer_read %w10[%c0, %c3], %cst {in_bounds=[true]}: tensor<?x?xf32>, vector<2xf32>
%u01 = "test.some_use"(%r01) : (vector<2xf32>) -> vector<2xf32>
- %w11 = vector.transfer_write %u01, %w10[%c0, %c3] : vector<2xf32>, tensor<?x?xf32>
+ %w11 = vector.transfer_write %u01, %w10[%c0, %c3] {in_bounds=[true]} : vector<2xf32>, tensor<?x?xf32>
scf.yield %w11 : tensor<?x?xf32>
}
return %1 : tensor<?x?xf32>
More information about the Mlir-commits
mailing list