[mlir][tensor] Add tensor.dim operation

[springerm]

Summary

• Split memref.dim into two operations: memref.dim and tensor.dim. Both ops have the same builder interface and op argument names, so that they can be used with templates in patterns that apply to both tensors and memrefs (e.g., some patterns in Linalg).
• Add constant materializer to TensorDialect (needed for folding in affine.apply etc.).
• Remove some MemRefDialect dependencies, make some explicit.

[springerm]

small improvement

[springerm]

fix test

[ftynse]

mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td

88–89

mlir/lib/Dialect/Affine/IR/AffineOps.cpp

68

I'm not sure Affine is supposed to work on tensors. Does something break if you don't add this check?

mlir/lib/Dialect/MemRef/IR/MemRefDialect.cpp

10

This is rather unfortunate to have MemRef depend on Tensor only to provide one helper function. Isn't there a better place? It seems to be predominantly used in Linalg, can it live there (and the remaining users can dispatch themselves if necessary; like SparseTensor is unlikely to work on memrefs).

[springerm]

mlir/lib/Dialect/MemRef/IR/MemRefDialect.cpp

10

I can put the helper function somewhere else, but MemRef would still depend on Tensor. (This dependency is not new, I only added an include here.)

There are MemRef canonicalization patterns unrelated to this change, that create/use Tensor dialect ops: LoadOfBufferCast, BufferCast

To give you the full picture: createOrFoldDimOp is also used in places outside of Linalg:

• getDynOperands (memref dialect). However, this function could also go somewhere else.
• VectorTransforms.cpp because vector.transfer_read/write work on tensors and memrefs.
• VectorToSCF.cpp (same reason)
• An internal G3 project

Should I duplicate the function is those places or leave it here?

[springerm]

mlir/lib/Dialect/Affine/IR/AffineOps.cpp

68

This is actually needed. Otherwise test/Dialect/Linalg/fusion-sequence.mlir and others are failing. (The generated IR, in particular affine maps are slightly different.)

[silvas]

mlir/lib/Dialect/StandardOps/Transforms/Bufferize.cpp

26

This pattern can be moved to mlir/lib/Dialect/Tensor/Transforms/Bufferize.cpp (can be done in separate patch).

mlir/lib/Dialect/Tensor/Transforms/Bufferize.cpp

78

We shouldn't need this pattern.

We should already be handling this in mlir/lib/Dialect/StandardOps/Transforms/TensorConstantBufferize.cpp. Can you try adding that to your pipeline to see if it fixes things?

[silvas]

(before submitting, please wait for LGTM from at least one other person since this change covers a lot of stuff)

[ftynse]

mlir/lib/Dialect/MemRef/IR/MemRefDialect.cpp

10

Cross-dialect canonicalizations are annoying but okay, this commit is not adding the dependency, just making it deeper.

getDynOperands (memref dialect). However, this function could also go somewhere else.

Yes, and this commit adds the use of createOrFoldDimOp there.

An internal G3 project

It must adapt to whatever we decide to do upstream.

So the only other actual user is vector. Well, I won't block this refactoring on code placement decision. But let's add a todo to revisit this. I remember @silvas floated an idea of a separate "tensor-to-memref" dialect that could contain the interface stuff.

[aartbik]

mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp

380

Thanks for the rewrite. Does the use of createOrFold here mean tha tI can remove the special casing I added to deal with constant vs. unknown dim ops?

(I can do that in a follow up, but just curious)

[springerm]

mlir/lib/Dialect/MemRef/IR/MemRefDialect.cpp

10

Added TODO, will prepare a commit that moves the function once we decided where it should go.

mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp

380

This function creates a memref::DimOp or tensor::DimOp, depending on the the type of t->get(). Basically just forwards the arguments to the constructor, nothing more.

I'm not familiar with the code here, but if t->get() is always a tensor, then this should be replaced with rewriter.create<tensor::DimOp>(...).

[springerm]

address comments

mlir/include/mlir/Conversion/Passes.td

 // ShapeToStandard
 //===----------------------------------------------------------------------===//
 
 def ConvertShapeToStandard : Pass<"convert-shape-to-std", "ModuleOp"> {
   let summary = "Convert operations from the shape dialect into the standard "
                 "dialect";
   let constructor = "mlir::createConvertShapeToStandardPass()";
   let dependentDialects = [
-    "memref::MemRefDialect",
     "StandardOpsDialect",
     "scf::SCFDialect",
     "tensor::TensorDialect"
   ];
 }
 
 def ConvertShapeConstraints: Pass<"convert-shape-constraints", "FuncOp"> {
   let summary = "Convert shape constraint operations to the standard dialect";

mlir/include/mlir/Dialect/MemRef/IR/MemRef.h

 /// entry contains either the dynamic value or a ConstantIndexOp constructed
 /// with `b` at location `loc`.
 SmallVector<Range, 8> getOrCreateRanges(OffsetSizeAndStrideOpInterface op,
                                         OpBuilder &b, Location loc);
 
 /// Given an operation, retrieves the value of each dynamic dimension through
 /// constructing the necessary DimOp operators.
 SmallVector<Value, 4> getDynOperands(Location loc, Value val, OpBuilder &b);
+
+// Helper function that creates a memref::DimOp or tensor::DimOp depending on
+// the type of `source`.
+Value createOrFoldDimOp(OpBuilder &b, Location loc, Value source, int64_t dim);
 } // namespace mlir
 
 //===----------------------------------------------------------------------===//
 // MemRef Dialect
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/MemRef/IR/MemRefOpsDialect.h.inc"
 

mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td

def MemRef_DeallocOp : MemRef_Op<"dealloc", [MemRefsNormalizable]> {
   let verifier = ?;
   let assemblyFormat = "$memref attr-dict `:` type($memref)";
 }
 
 //===----------------------------------------------------------------------===//
 // DimOp
 //===----------------------------------------------------------------------===//
 
-def DimOp : MemRef_Op<"dim", [NoSideEffect, MemRefsNormalizable]> {
+def MemRef_DimOp : MemRef_Op<"dim", [NoSideEffect, MemRefsNormalizable]> {
   let summary = "dimension index operation";
   let description = [{
     The `dim` operation takes a memref and a dimension operand of type `index`.
     It returns the size of the requested dimension of the given memref.
     If the dimension index is out of bounds the behavior is undefined.
 
     The specified memref type is that of the first operand.
 
let description = [{
     %y = memref.dim %A, %c1 : memref<4 x ? x f32>
 
     // Equivalent generic form:
     %x = "memref.dim"(%A, %c0) : (memref<4 x ? x f32>, index) -> index
     %y = "memref.dim"(%A, %c1) : (memref<4 x ? x f32>, index) -> index
     ```
   }];
 
-  let arguments = (ins AnyTypeOf<[AnyTensor, AnyRankedOrUnrankedMemRef],
-                                 "any memref or tensor type">:$memrefOrTensor,
+  let arguments = (ins AnyRankedOrUnrankedMemRef:$source,
                        Index:$index);
   let results = (outs Index:$result);
 
   let assemblyFormat = [{
-    attr-dict $memrefOrTensor `,` $index `:` type($memrefOrTensor)
+    attr-dict $source `,` $index `:` type($source)
   }];
 
   let builders = [
-    OpBuilder<(ins "Value":$memrefOrTensor, "int64_t":$index)>,
-    OpBuilder<(ins "Value":$memrefOrTensor, "Value":$index)>
+    OpBuilder<(ins "Value":$source, "int64_t":$index)>,
+    OpBuilder<(ins "Value":$source, "Value":$index)>
   ];
 
   let extraClassDeclaration = [{
     /// Helper function to get the index as a simple integer if it is constant.
     Optional<int64_t> getConstantIndex();
   }];
 
   let hasCanonicalizer = 1;
TensorType getType() {
       if (resultType.isa<TensorType>())
         return resultType.cast<TensorType>();
       return {};
     }
   }];
 
   let assemblyFormat = "$memref attr-dict `:` type($memref)";
 
+  let hasCanonicalizer = 1;
   let hasFolder = 1;
 }
 
 //===----------------------------------------------------------------------===//
 // TensorStoreOp
 //===----------------------------------------------------------------------===//
 
 def TensorStoreOp : MemRef_Op<"tensor_store",

mlir/include/mlir/Dialect/StandardOps/Transforms/Passes.td

 //===-- Passes.td - StandardOps pass definition file -------*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef MLIR_DIALECT_STANDARD_TRANSFORMS_PASSES
 #define MLIR_DIALECT_STANDARD_TRANSFORMS_PASSES
 
 include "mlir/Pass/PassBase.td"
 
 def StdBufferize : FunctionPass<"std-bufferize"> {
   let summary = "Bufferize the std dialect";
   let constructor = "mlir::createStdBufferizePass()";
-  let dependentDialects = ["scf::SCFDialect"];
+  let dependentDialects = ["memref::MemRefDialect", "scf::SCFDialect"];
 }
 
 def StdExpandOps : FunctionPass<"std-expand"> {
   let summary = "Legalize std operations to be convertible to LLVM.";
   let constructor = "mlir::createStdExpandOpsPass()";
 }
 
 def FuncBufferize : Pass<"func-bufferize", "ModuleOp"> {

mlir/include/mlir/Dialect/Tensor/IR/TensorBase.td

 //===- TensorBase.td - Base definitions for tensor dialect -*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef TENSOR_BASE
 #define TENSOR_BASE
 
 include "mlir/IR/OpBase.td"
 
 def Tensor_Dialect : Dialect {
   let name = "tensor";
   let cppNamespace = "::mlir::tensor";
+
   let description = [{
     The `tensor` dialect is intended to hold core tensor creation and
     manipulation ops, which are not strongly associated with any particular
     other dialect or domain abstraction. The primary smoke test of this is ops
     that make sense for any tensor element type.
 
     We leave it to other dialects to hold the vast swath of possible
     computations one might want to do on a tensor.
let description = [{
     scope as much as possible. The expectation is that at some point
     in the future, the `tensor` dialect’s scope may be broadened through a
     careful discussion of the tradeoffs.
 
     The `tensor` type is actually a builtin type (it lives in the builtin
     dialect), and does not live in this dialect.
 
   }];
+
+  let hasConstantMaterializer = 1;
 }
 
 #endif // TENSOR_BASE

mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td

def Tensor_CastOp : Tensor_Op<"cast", [
   let results = (outs AnyTensor:$dest);
   let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
 
   let hasCanonicalizer = 1;
   let verifier = ?;
 }
 
 //===----------------------------------------------------------------------===//
+// DimOp
+//===----------------------------------------------------------------------===//
+
+def Tensor_DimOp : Tensor_Op<"dim", [NoSideEffect]> {
+  let summary = "dimension index operation";
+  let description = [{
+    The `dim` operation takes a tensor and a dimension operand of type `index`.
+    It returns the size of the requested dimension of the given tensor.
+    If the dimension index is out of bounds, the behavior is undefined.
+
+    The specified tensor type is that of the first operand.
+
+    Example:
+
+    ```mlir
+    // Always returns 4, can be constant folded:
+    %c0 = constant 0 : index
+    %x = tensor.dim %A, %c0 : tensor<4x?xf32>
+
+    // Returns the dynamic dimension of %A.
+    %c1 = constant 1 : index
+    %y = tensor.dim %A, %c1 : memref<4x?xf32>
+
+    // Equivalent generic form:
+    %x = "tensor.dim"(%A, %c0) : (memref<4x?xf32>, index) -> index
+    %y = "tensor.dim"(%A, %c1) : (memref<4x?xf32>, index) -> index
+    ```
+  }];
+
+  let arguments = (ins AnyTensor:$source,
+                       Index:$index);
+  let results = (outs Index:$result);
+
+  let assemblyFormat = [{
+    attr-dict $source `,` $index `:` type($source)
+  }];
+
+  let builders = [
+    OpBuilder<(ins "Value":$source, "int64_t":$index)>,
+    OpBuilder<(ins "Value":$source, "Value":$index)>
+  ];
+
+  let extraClassDeclaration = [{
+    /// Helper function to get the index as a simple integer if it is constant.
+    Optional<int64_t> getConstantIndex();
+  }];
+
+  let hasCanonicalizer = 1;
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
 // ExtractOp
 //===----------------------------------------------------------------------===//
 
 def Tensor_ExtractOp : Tensor_Op<"extract",
     [NoSideEffect,
      TypesMatchWith<"result type matches element type of tensor",
                     "tensor", "result",
                     "$_self.cast<ShapedType>().getElementType()">]> {

mlir/include/mlir/Dialect/Tensor/Transforms/Passes.td

 //===-- Passes.td - pass definition file -------------------*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef MLIR_DIALECT_TENSOR_TRANSFORMS_PASSES
 #define MLIR_DIALECT_TENSOR_TRANSFORMS_PASSES
 
 include "mlir/Pass/PassBase.td"
 
 def TensorBufferize : FunctionPass<"tensor-bufferize"> {
   let summary = "Bufferize the `tensor` dialect";
   let constructor = "mlir::createTensorBufferizePass()";
-  let dependentDialects = ["scf::SCFDialect"];
+  let dependentDialects = ["scf::SCFDialect", "memref::MemRefDialect"];
 }
 
 #endif // MLIR_DIALECT_TENSOR_TRANSFORMS_PASSES

mlir/lib/Conversion/ShapeToStandard/CMakeLists.txt

add_mlir_conversion_library(MLIRShapeToStandard
   MLIRConversionPassIncGen
   ShapeToStandardIncGen
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRIR
-  MLIRMemRef
   MLIRShape
   MLIRTensor
   MLIRPass
   MLIRSCF
   MLIRTransforms
   )

mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp

 //===- ShapeToStandard.cpp - conversion from Shape to Standard dialect ----===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Conversion/ShapeToStandard/ShapeToStandard.h"
 
 #include "../PassDetail.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/SCF/SCF.h"
 #include "mlir/Dialect/Shape/IR/Shape.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/BlockAndValueMapping.h"
 #include "mlir/IR/ImplicitLocOpBuilder.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "llvm/ADT/STLExtras.h"
LogicalResult BroadcastOpConverter::matchAndRewrite(
   Value zero = lb.create<ConstantIndexOp>(0);
   Type indexTy = lb.getIndexType();
 
   // Save all the ranks for bounds checking. Because this is a tensor
   // representing the shape extents, the rank is the extent of the only
   // dimension in the tensor.
   SmallVector<Value> ranks, rankDiffs;
   llvm::append_range(ranks, llvm::map_range(transformed.shapes(), [&](Value v) {
-                       return lb.create<memref::DimOp>(v, zero);
+                       return lb.create<tensor::DimOp>(v, zero);
                      }));
 
   // Find the maximum rank
   Value maxRank = ranks.front();
   for (Value v : llvm::drop_begin(ranks, 1)) {
     Value rankIsGreater = lb.create<CmpIOp>(CmpIPredicate::ugt, v, maxRank);
     maxRank = lb.create<SelectOp>(rankIsGreater, v, maxRank);
   }
LogicalResult IsBroadcastableOpConverter::matchAndRewrite(
   Value one = lb.create<ConstantIndexOp>(1);
   Type indexTy = lb.getIndexType();
 
   // Save all the ranks for bounds checking. Because this is a tensor
   // representing the shape extents, the rank is the extent of the only
   // dimension in the tensor.
   SmallVector<Value> ranks, rankDiffs;
   llvm::append_range(ranks, llvm::map_range(transformed.shapes(), [&](Value v) {
-                       return lb.create<memref::DimOp>(v, zero);
+                       return lb.create<tensor::DimOp>(v, zero);
                      }));
 
   // Find the maximum rank
   Value maxRank = ranks.front();
   for (Value v : llvm::drop_begin(ranks, 1)) {
     Value rankIsGreater = lb.create<CmpIOp>(CmpIPredicate::ugt, v, maxRank);
     maxRank = lb.create<SelectOp>(rankIsGreater, v, maxRank);
   }
LogicalResult GetExtentOpConverter::matchAndRewrite(
   // For now, only error-free types are supported by this lowering.
   if (op.getType().isa<SizeType>())
     return failure();
 
   // Derive shape extent directly from shape origin if possible. This
   // circumvents the necessity to materialize the shape in memory.
   if (auto shapeOfOp = op.shape().getDefiningOp<ShapeOfOp>()) {
     if (shapeOfOp.arg().getType().isa<ShapedType>()) {
-      rewriter.replaceOpWithNewOp<memref::DimOp>(op, shapeOfOp.arg(),
+      rewriter.replaceOpWithNewOp<tensor::DimOp>(op, shapeOfOp.arg(),
                                                  transformed.dim());
       return success();
     }
   }
 
   rewriter.replaceOpWithNewOp<tensor::ExtractOp>(op, rewriter.getIndexType(),
                                                  transformed.shape(),
                                                  ValueRange{transformed.dim()});

 LogicalResult
 RankOpConverter::matchAndRewrite(shape::RankOp op, ArrayRef<Value> operands,
                                  ConversionPatternRewriter &rewriter) const {
   // For now, this lowering supports only error-free types.
   if (op.getType().isa<SizeType>())
     return failure();
 
   shape::RankOp::Adaptor transformed(operands);
-  rewriter.replaceOpWithNewOp<memref::DimOp>(op, transformed.shape(), 0);
+  rewriter.replaceOpWithNewOp<tensor::DimOp>(op, transformed.shape(), 0);
   return success();
 }
 
 namespace {
 /// Converts `shape.reduce` to `scf.for`.
 struct ReduceOpConverter : public OpConversionPattern<shape::ReduceOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
ReduceOpConverter::matchAndRewrite(shape::ReduceOp op, ArrayRef<Value> operands,
 
   auto loc = op.getLoc();
   shape::ReduceOp::Adaptor transformed(operands);
 
   Value zero = rewriter.create<ConstantIndexOp>(loc, 0);
   Value one = rewriter.create<ConstantIndexOp>(loc, 1);
   Type indexTy = rewriter.getIndexType();
   Value rank =
-      rewriter.create<memref::DimOp>(loc, indexTy, transformed.shape(), zero);
+      rewriter.create<tensor::DimOp>(loc, indexTy, transformed.shape(), zero);
 
   auto loop = rewriter.create<scf::ForOp>(
       loc, zero, rank, one, op.initVals(),
       [&](OpBuilder &b, Location loc, Value iv, ValueRange args) {
         Value extent =
             b.create<tensor::ExtractOp>(loc, transformed.shape(), iv);
 
         SmallVector<Value, 2> mappedValues{iv, extent};
ShapeEqOpConverter::matchAndRewrite(ShapeEqOp op, ArrayRef<Value> operands,
   }
 
   ShapeEqOp::Adaptor transformed(operands);
   auto loc = op.getLoc();
   Type indexTy = rewriter.getIndexType();
   Value zero = rewriter.create<ConstantIndexOp>(loc, 0);
   Value firstShape = transformed.shapes().front();
   Value firstRank =
-      rewriter.create<memref::DimOp>(loc, indexTy, firstShape, zero);
+      rewriter.create<tensor::DimOp>(loc, indexTy, firstShape, zero);
   Value result = nullptr;
   // Generate a linear sequence of compares, all with firstShape as lhs.
   for (Value shape : transformed.shapes().drop_front(1)) {
-    Value rank = rewriter.create<memref::DimOp>(loc, indexTy, shape, zero);
+    Value rank = rewriter.create<tensor::DimOp>(loc, indexTy, shape, zero);
     Value eqRank =
         rewriter.create<CmpIOp>(loc, CmpIPredicate::eq, firstRank, rank);
     auto same = rewriter.create<IfOp>(
         loc, i1Ty, eqRank,
         [&](OpBuilder &b, Location loc) {
           Value one = b.create<ConstantIndexOp>(loc, 1);
           Value init = b.create<ConstantOp>(loc, i1Ty, b.getBoolAttr(true));
           auto loop = b.create<scf::ForOp>(
LogicalResult ShapeOfOpConversion::matchAndRewrite(
   if (tensorTy.isa<RankedTensorType>()) {
 
     // Build values for individual extents.
     SmallVector<Value, 8> extentValues;
     RankedTensorType rankedTensorTy = tensorTy.cast<RankedTensorType>();
     int64_t rank = rankedTensorTy.getRank();
     for (int64_t i = 0; i < rank; i++) {
       if (rankedTensorTy.isDynamicDim(i)) {
-        Value extent = rewriter.create<memref::DimOp>(loc, tensor, i);
+        Value extent = rewriter.create<tensor::DimOp>(loc, tensor, i);
         extentValues.push_back(extent);
       } else {
         Value extent =
             rewriter.create<ConstantIndexOp>(loc, rankedTensorTy.getDimSize(i));
         extentValues.push_back(extent);
       }
     }
 
     // Materialize extent tensor.
     Value staticExtentTensor = rewriter.create<tensor::FromElementsOp>(
         loc, rewriter.getIndexType(), extentValues);
     rewriter.replaceOpWithNewOp<tensor::CastOp>(op, op.getType(),
                                                 staticExtentTensor);
     return success();
   }
 
   // Lower to `tensor.generate` otherwise.
   auto *ctx = rewriter.getContext();
   Value rank = rewriter.create<mlir::RankOp>(loc, tensor);
   rewriter.replaceOpWithNewOp<tensor::GenerateOp>(
       op, getExtentTensorType(ctx), ValueRange{rank},
       [&](OpBuilder &b, Location loc, ValueRange args) {
         Value dim = args.front();
-        Value extent = b.create<memref::DimOp>(loc, tensor, dim);
+        Value extent = b.create<tensor::DimOp>(loc, tensor, dim);
         b.create<tensor::YieldOp>(loc, extent);
       });
 
   return success();
 }
 
 namespace {
 class SplitAtOpConversion : public OpConversionPattern<SplitAtOp> {
LogicalResult SplitAtOpConversion::matchAndRewrite(
   // results are extent tensors.
   if (llvm::any_of(ValueRange{op.operand(), op.head(), op.tail()},
                    [](Value v) { return v.getType().isa<ShapeType>(); }))
     return failure();
 
   SplitAtOp::Adaptor transformed(op);
   ImplicitLocOpBuilder b(op.getLoc(), rewriter);
   Value zero = b.create<ConstantIndexOp>(0);
-  Value rank = b.create<memref::DimOp>(transformed.operand(), zero);
+  Value rank = b.create<tensor::DimOp>(transformed.operand(), zero);
 
   // index < 0 ? index + rank : index
   Value originalIndex = transformed.index();
   Value add = b.create<AddIOp>(originalIndex, rank);
   Value indexIsNegative =
       b.create<CmpIOp>(CmpIPredicate::slt, originalIndex, zero);
   Value index = b.create<SelectOp>(indexIsNegative, add, originalIndex);
 
class ConvertShapeToStandardPass
   void runOnOperation() override;
 };
 } // namespace
 
 void ConvertShapeToStandardPass::runOnOperation() {
   // Setup target legality.
   MLIRContext &ctx = getContext();
   ConversionTarget target(ctx);
-  target.addLegalDialect<memref::MemRefDialect, StandardOpsDialect, SCFDialect,
-                         tensor::TensorDialect>();
+  target
+      .addLegalDialect<StandardOpsDialect, SCFDialect, tensor::TensorDialect>();
   target.addLegalOp<CstrRequireOp, FuncOp, ModuleOp>();
 
   // Setup conversion patterns.
   RewritePatternSet patterns(&ctx);
   populateShapeToStandardConversionPatterns(patterns);
 
   // Apply conversion.
   auto module = getOperation();

mlir/lib/Conversion/StandardToLLVM/StandardToLLVM.cpp

 // A `dim` is converted to a constant for static sizes and to an access to the
 // size stored in the memref descriptor for dynamic sizes.
 struct DimOpLowering : public ConvertOpToLLVMPattern<memref::DimOp> {
   using ConvertOpToLLVMPattern<memref::DimOp>::ConvertOpToLLVMPattern;
 
   LogicalResult
   matchAndRewrite(memref::DimOp dimOp, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
-    Type operandType = dimOp.memrefOrTensor().getType();
+    Type operandType = dimOp.source().getType();
     if (operandType.isa<UnrankedMemRefType>()) {
       rewriter.replaceOp(dimOp, {extractSizeOfUnrankedMemRef(
                                     operandType, dimOp, operands, rewriter)});
 
       return success();
     }
     if (operandType.isa<MemRefType>()) {
       rewriter.replaceOp(dimOp, {extractSizeOfRankedMemRef(
                                     operandType, dimOp, operands, rewriter)});
       return success();
     }
-    return failure();
+    llvm_unreachable("expected MemRefType or UnrankedMemRefType");
   }
 
 private:
   Value extractSizeOfUnrankedMemRef(Type operandType, memref::DimOp dimOp,
                                     ArrayRef<Value> operands,
                                     ConversionPatternRewriter &rewriter) const {
     Location loc = dimOp.getLoc();
     memref::DimOp::Adaptor transformed(operands);
 
     auto unrankedMemRefType = operandType.cast<UnrankedMemRefType>();
     auto scalarMemRefType =
         MemRefType::get({}, unrankedMemRefType.getElementType());
     unsigned addressSpace = unrankedMemRefType.getMemorySpaceAsInt();
 
     // Extract pointer to the underlying ranked descriptor and bitcast it to a
     // memref<element_type> descriptor pointer to minimize the number of GEP
     // operations.
-    UnrankedMemRefDescriptor unrankedDesc(transformed.memrefOrTensor());
+    UnrankedMemRefDescriptor unrankedDesc(transformed.source());
     Value underlyingRankedDesc = unrankedDesc.memRefDescPtr(rewriter, loc);
     Value scalarMemRefDescPtr = rewriter.create<LLVM::BitcastOp>(
         loc,
         LLVM::LLVMPointerType::get(typeConverter->convertType(scalarMemRefType),
                                    addressSpace),
         underlyingRankedDesc);
 
     // Get pointer to offset field of memref<element_type> descriptor.
Value extractSizeOfRankedMemRef(Type operandType, memref::DimOp dimOp,
     Location loc = dimOp.getLoc();
     memref::DimOp::Adaptor transformed(operands);
     // Take advantage if index is constant.
     MemRefType memRefType = operandType.cast<MemRefType>();
     if (Optional<int64_t> index = dimOp.getConstantIndex()) {
       int64_t i = index.getValue();
       if (memRefType.isDynamicDim(i)) {
         // extract dynamic size from the memref descriptor.
-        MemRefDescriptor descriptor(transformed.memrefOrTensor());
+        MemRefDescriptor descriptor(transformed.source());
         return descriptor.size(rewriter, loc, i);
       }
       // Use constant for static size.
       int64_t dimSize = memRefType.getDimSize(i);
       return createIndexConstant(rewriter, loc, dimSize);
     }
     Value index = dimOp.index();
     int64_t rank = memRefType.getRank();
-    MemRefDescriptor memrefDescriptor(transformed.memrefOrTensor());
+    MemRefDescriptor memrefDescriptor(transformed.source());
     return memrefDescriptor.size(rewriter, loc, index, rank);
   }
 };
 
 struct RankOpLowering : public ConvertOpToLLVMPattern<RankOp> {
   using ConvertOpToLLVMPattern<RankOp>::ConvertOpToLLVMPattern;
 
   LogicalResult

mlir/lib/Conversion/TosaToLinalg/CMakeLists.txt

 add_mlir_conversion_library(MLIRTosaToLinalg
   TosaToLinalg.cpp
   TosaToLinalgPass.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Tosa
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/IR
 
   DEPENDS
   MLIRConversionPassIncGen
 
   LINK_LIBS PUBLIC
   MLIRIR
   MLIRLinalg
   MLIRLinalgUtils
   MLIRMath
-  MLIRMemRef
   MLIRPass
   MLIRTensor
   MLIRTosa
   MLIRTosaTransforms
   MLIRSupport
   )

mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp

 //===- TosaToLinalg.cpp - Lowering Tosa to Linalg Dialect -----------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // These rewriters lower from the Tosa to the Linalg dialect.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Conversion/TosaToLinalg/TosaToLinalg.h"
 #include "mlir/Dialect/Linalg/IR/LinalgOps.h"
 #include "mlir/Dialect/Math/IR/Math.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/IR/Matchers.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
Value axisValue =
         rewriter.create<ConstantOp>(loc, rewriter.getIndexAttr(axis));
     int rank = resultType.getRank();
     SmallVector<Value, 3> offsets, sizes, strides;
     sizes.reserve(rank);
     strides.resize(rank, rewriter.create<ConstantIndexOp>(loc, 1));
     offsets.resize(rank, rewriter.create<ConstantIndexOp>(loc, 0));
 
     for (int i = 0; i < rank; ++i) {
-      sizes.push_back(rewriter.create<memref::DimOp>(loc, args[0], i));
+      sizes.push_back(rewriter.create<tensor::DimOp>(loc, args[0], i));
     }
 
     Value resultDimSize = sizes[axis];
     for (auto arg : args.drop_front()) {
-      auto size = rewriter.create<memref::DimOp>(loc, arg, axisValue);
+      auto size = rewriter.create<tensor::DimOp>(loc, arg, axisValue);
       resultDimSize = rewriter.create<AddIOp>(loc, resultDimSize, size);
     }
     sizes[axis] = resultDimSize;
 
     Value init = rewriter.create<linalg::InitTensorOp>(
         loc, resultType.getShape(), resultType.getElementType());
 
     Value zeroVal = rewriter.create<ConstantOp>(
         loc, rewriter.getZeroAttr(resultType.getElementType()));
     Value result =
         rewriter.create<linalg::FillOp>(loc, zeroVal, init).getResult(0);
 
     for (auto arg : args) {
-      sizes[axis] = rewriter.create<memref::DimOp>(loc, arg, axisValue);
+      sizes[axis] = rewriter.create<tensor::DimOp>(loc, arg, axisValue);
       result = rewriter.create<tensor::InsertSliceOp>(loc, arg, result, offsets,
                                                       sizes, strides);
       offsets[axis] = rewriter.create<AddIOp>(loc, offsets[axis], sizes[axis]);
     }
     rewriter.replaceOp(op, result);
     return success();
   }
 };

mlir/lib/Conversion/TosaToLinalg/TosaToLinalgPass.cpp

 //===- TosaToLinalgPass.cpp - Lowering Tosa to Linalg Dialect -------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This transformation pass legalizes Tosa operations to the Linalg dialect.
 //
 //===----------------------------------------------------------------------===//
 
 #include "../PassDetail.h"
 #include "mlir/Conversion/TosaToLinalg/TosaToLinalg.h"
 #include "mlir/Dialect/Linalg/IR/LinalgOps.h"
 #include "mlir/Dialect/Math/IR/Math.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/Dialect/Tosa/Transforms/PassDetail.h"
 #include "mlir/Dialect/Tosa/Transforms/Passes.h"
 #include "mlir/Dialect/Tosa/Utils/QuantUtils.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/Pass/PassManager.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
 using namespace mlir;
 
 namespace {
 struct TosaToLinalgOnTensors
     : public TosaToLinalgOnTensorsBase<TosaToLinalgOnTensors> {
 public:
   void getDependentDialects(DialectRegistry &registry) const override {
     registry.insert<linalg::LinalgDialect, math::MathDialect,
-                    memref::MemRefDialect, StandardOpsDialect,
-                    tensor::TensorDialect>();
+                    StandardOpsDialect, tensor::TensorDialect>();
   }
 
   void runOnFunction() override {
     RewritePatternSet patterns(&getContext());
     ConversionTarget target(getContext());
-    target.addLegalDialect<linalg::LinalgDialect, memref::MemRefDialect,
-                           StandardOpsDialect, tensor::TensorDialect>();
+    target.addLegalDialect<linalg::LinalgDialect, StandardOpsDialect,
+                           tensor::TensorDialect>();
     target.addIllegalDialect<tosa::TosaDialect>();
 
     // Not every TOSA op can be legalized to linalg.
     target.addLegalOp<tosa::ApplyScaleOp>();
     target.addLegalOp<tosa::IfOp>();
     target.addLegalOp<tosa::ConstOp>();
     target.addLegalOp<tosa::WhileOp>();
 

mlir/lib/Conversion/VectorToSCF/VectorToSCF.cpp

static Value generateInBoundsCheck(
   bool hasRetVal = !resultTypes.empty();
   Value cond; // Condition to be built...
 
   // Condition check 1: Access in-bounds?
   bool isBroadcast = !dim.hasValue(); // No in-bounds check for broadcasts.
   Location loc = xferOp.getLoc();
   ImplicitLocOpBuilder lb(xferOp.getLoc(), b);
   if (!xferOp.isDimInBounds(0) && !isBroadcast) {
-    Value memrefDim = lb.create<memref::DimOp>(xferOp.source(), *dim);
+    Value memrefDim = createOrFoldDimOp(b, loc, xferOp.source(), *dim);
     AffineExpr d0, d1;
     bindDims(xferOp.getContext(), d0, d1);
     Value base = xferOp.indices()[dim.getValue()];
     Value memrefIdx = makeComposedAffineApply(b, loc, d0 + d1, {base, iv});
     cond = lb.create<CmpIOp>(CmpIPredicate::sgt, memrefDim, memrefIdx);
   }
 
   // Condition check 2: Masked in?

mlir/lib/Dialect/Affine/IR/AffineOps.cpp

if (!isTopLevelValue(value, src))
     return true;
 
   // If it's a top-level value because it's a block operand, i.e. a
   // function argument, check whether the value replacing it after
   // inlining is a valid dimension in the new region.
   if (value.isa<BlockArgument>())
     return legalityCheck(mapping.lookup(value), dest);
 
-  // If it's a top-level value beacuse it's defined in the region,
+  // If it's a top-level value because it's defined in the region,
   // it can only be inlined if the defining op is a constant or a
   // `dim`, which can appear anywhere and be valid, since the defining
   // op won't be top-level anymore after inlining.
   Attribute operandCst;
   return matchPattern(value.getDefiningOp(), m_Constant(&operandCst)) ||
-         value.getDefiningOp<memref::DimOp>();
+         value.getDefiningOp<memref::DimOp>() ||
+         value.getDefiningOp<tensor::DimOp>();
 }
 
 /// Checks if all values known to be legal affine dimensions or symbols in `src`
 /// remain so if their respective users are inlined into `dest`.
 static bool
 remainsLegalAfterInline(ValueRange values, Region *src, Region *dest,
                         const BlockAndValueMapping &mapping,
                         function_ref<bool(Value, Region *)> legalityCheck) {
bool mlir::isValidDim(Value value, Region *region) {
   }
 
   // Affine apply operation is ok if all of its operands are ok.
   if (auto applyOp = dyn_cast<AffineApplyOp>(op))
     return applyOp.isValidDim(region);
   // The dim op is okay if its operand memref/tensor is defined at the top
   // level.
   if (auto dimOp = dyn_cast<memref::DimOp>(op))
-    return isTopLevelValue(dimOp.memrefOrTensor());
+    return isTopLevelValue(dimOp.source());
+  if (auto dimOp = dyn_cast<tensor::DimOp>(op))
+    return isTopLevelValue(dimOp.source());
   return false;
 }
 
 /// Returns true if the 'index' dimension of the `memref` defined by
 /// `memrefDefOp` is a statically  shaped one or defined using a valid symbol
 /// for `region`.
 template <typename AnyMemRefDefOp>
 static bool isMemRefSizeValidSymbol(AnyMemRefDefOp memrefDefOp, unsigned index,
                                     Region *region) {
   auto memRefType = memrefDefOp.getType();
   // Statically shaped.
   if (!memRefType.isDynamicDim(index))
     return true;
   // Get the position of the dimension among dynamic dimensions;
   unsigned dynamicDimPos = memRefType.getDynamicDimIndex(index);
   return isValidSymbol(*(memrefDefOp.getDynamicSizes().begin() + dynamicDimPos),
                        region);
 }
 
 /// Returns true if the result of the dim op is a valid symbol for `region`.
-static bool isDimOpValidSymbol(memref::DimOp dimOp, Region *region) {
-  // The dim op is okay if its operand memref is defined at the top level.
-  if (isTopLevelValue(dimOp.memrefOrTensor()))
+template <typename OpTy>
+static bool isDimOpValidSymbol(OpTy dimOp, Region *region) {
+  // The dim op is okay if its source is defined at the top level.
+  if (isTopLevelValue(dimOp.source()))
     return true;
 
   // Conservatively handle remaining BlockArguments as non-valid symbols.
   // E.g. scf.for iterArgs.
-  if (dimOp.memrefOrTensor().isa<BlockArgument>())
+  if (dimOp.source().template isa<BlockArgument>())
     return false;
 
   // The dim op is also okay if its operand memref is a view/subview whose
   // corresponding size is a valid symbol.
   Optional<int64_t> index = dimOp.getConstantIndex();
   assert(index.hasValue() &&
          "expect only `dim` operations with a constant index");
   int64_t i = index.getValue();
-  return TypeSwitch<Operation *, bool>(dimOp.memrefOrTensor().getDefiningOp())
+  return TypeSwitch<Operation *, bool>(dimOp.source().getDefiningOp())
       .Case<memref::ViewOp, memref::SubViewOp, memref::AllocOp>(
           [&](auto op) { return isMemRefSizeValidSymbol(op, i, region); })
       .Default([](Operation *) { return false; });
 }
 
 // A value can be used as a symbol (at all its use sites) iff it meets one of
 // the following conditions:
 // *) It is a constant.
if (isTopLevelValue(value))
     return true;
 
   if (auto *defOp = value.getDefiningOp())
     return isValidSymbol(value, getAffineScope(defOp));
 
   return false;
 }
 
-/// A value can be used as a symbol for `region` iff it meets onf of the the
+/// A value can be used as a symbol for `region` iff it meets one of the
 /// following conditions:
 /// *) It is a constant.
 /// *) It is the result of an affine apply operation with symbol arguments.
 /// *) It is a result of the dim op on a memref whose corresponding size is
 ///    a valid symbol.
 /// *) It is defined at the top level of 'region' or is its argument.
 /// *) It dominates `region`'s parent op.
 /// If `region` is null, conservatively assume the symbol definition scope does
bool mlir::isValidSymbol(Value value, Region *region) {
 
   // Affine apply operation is ok if all of its operands are ok.
   if (auto applyOp = dyn_cast<AffineApplyOp>(defOp))
     return applyOp.isValidSymbol(region);
 
   // Dim op results could be valid symbols at any level.
   if (auto dimOp = dyn_cast<memref::DimOp>(defOp))
     return isDimOpValidSymbol(dimOp, region);
+  if (auto dimOp = dyn_cast<tensor::DimOp>(defOp))
+    return isDimOpValidSymbol(dimOp, region);
 
   // Check for values dominating `region`'s parent op.
   Operation *regionOp = region ? region->getParentOp() : nullptr;
   if (regionOp && !regionOp->hasTrait<OpTrait::IsIsolatedFromAbove>())
     if (auto *parentRegion = region->getParentOp()->getParentRegion())
       return isValidSymbol(value, parentRegion);
 
   return false;

mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp

SmallVector<Value, 4> mlir::linalg::applyMapToValues(OpBuilder &b, Location loc,
   return res;
 }
 
 SmallVector<Value, 4> LinalgOp::createFlatListOfOperandDims(OpBuilder &b,
                                                             Location loc) {
   SmallVector<Value, 4> res;
   for (OpOperand *opOperand : getInputAndOutputOperands()) {
     for (int64_t i = 0, e = getRank(opOperand); i < e; ++i)
-      res.push_back(b.createOrFold<memref::DimOp>(loc, opOperand->get(), i));
+      res.push_back(createOrFoldDimOp(b, loc, opOperand->get(), i));
   }
   return res;
 }
 
 SmallVector<int64_t, 4> LinalgOp::createFlatListOfOperandStaticDims() {
   SmallVector<int64_t, 4> res;
   assert(!hasDynamicShape() && "expected operands to have static shapes");
   for (OpOperand *opOperand : getInputAndOutputOperands())
auto allResultDimValues =
       applyMapToValues(b, loc, resultShapesFromInputShapesMap,
                        createFlatListOfOperandDims(b, loc));
   int64_t pos = 0;
   ArrayRef<AffineExpr> shapeExprs = resultShapesFromInputShapesMap.getResults();
   for (OpOperand *opOperand : getOutputOperands()) {
     SmallVector<Value> shapes;
     for (int64_t dim : llvm::seq<int64_t>(0, getRank(opOperand))) {
       if (checkDimExpr.visit(shapeExprs[pos]))
-        shapes.push_back(
-            b.createOrFold<memref::DimOp>(loc, opOperand->get(), dim));
+        shapes.push_back(createOrFoldDimOp(b, loc, opOperand->get(), dim));
       else
         shapes.push_back(allResultDimValues[pos]);
       pos++;
     }
     reifiedReturnShapes.emplace_back(std::move(shapes));
   }
   return success();
 }

mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp

 
 PadTensorOp PadTensorOp::createPadHighOp(Type type, Value source, Value pad,
                                          Location loc, OpBuilder &builder) {
   SmallVector<OpFoldResult, 4> low, high;
   auto rankedTensorType = type.cast<RankedTensorType>();
   assert(rankedTensorType.hasStaticShape());
   int rank = rankedTensorType.getRank();
   for (int i = 0; i < rank; ++i) {
-    auto dimOp = builder.createOrFold<memref::DimOp>(loc, source, i);
+    auto dimOp = builder.createOrFold<tensor::DimOp>(loc, source, i);
     auto resultDimSize = builder.createOrFold<ConstantIndexOp>(
         loc, rankedTensorType.getDimSize(i));
     auto highValue = builder.createOrFold<SubIOp>(loc, resultDimSize, dimOp);
     high.push_back(highValue);
     low.push_back(builder.createOrFold<ConstantIndexOp>(loc, 0));
   }
   return PadTensorOp::createPadScalarOp(type, source, pad, low, high, loc,
                                         builder);
 }
 
 LogicalResult PadTensorOp::reifyReturnTypeShapesPerResultDim(
     OpBuilder &b, SmallVectorImpl<SmallVector<Value>> &reifiedReturnShapes) {
   Location loc = getLoc();
   auto lowPad = getMixedLowPad();
   auto highPad = getMixedHighPad();
   SmallVector<Value> shapes;
   for (auto dim : llvm::seq<int64_t>(0, getSourceType().getRank())) {
     // Shape along each dimension is source dim + low pad + high pad.
     SmallVector<Value> mapOperands;
-    mapOperands.push_back(b.createOrFold<memref::DimOp>(loc, source(), dim));
+    mapOperands.push_back(b.createOrFold<tensor::DimOp>(loc, source(), dim));
     AffineExpr expr = b.getAffineDimExpr(0);
     unsigned numSymbols = 0;
     auto addOpFoldResult = [&](OpFoldResult valueOrAttr) {
       if (Value v = valueOrAttr.dyn_cast<Value>()) {
         expr = expr + b.getAffineSymbolExpr(numSymbols++);
         mapOperands.push_back(v);
         return;
       }
getCollapsedOutputDimFromInputShape(OpBuilder &builder, Location loc,
                                     ArrayRef<AffineMap> reassociationMap) {
   AffineMap map = reassociationMap[dimIndex];
   unsigned startPos =
       map.getResults().front().cast<AffineDimExpr>().getPosition();
   unsigned endPos = map.getResults().back().cast<AffineDimExpr>().getPosition();
   AffineExpr expr;
   SmallVector<Value, 2> dynamicDims;
   for (auto dim : llvm::seq(startPos, endPos + 1)) {
-    dynamicDims.push_back(builder.createOrFold<memref::DimOp>(loc, src, dim));
+    dynamicDims.push_back(builder.createOrFold<tensor::DimOp>(loc, src, dim));
     AffineExpr currExpr = builder.getAffineSymbolExpr(dim - startPos);
     expr = (expr ? expr * currExpr : currExpr);
   }
   return applyMapToValues(builder, loc,
                           AffineMap::get(0, endPos - startPos + 1, expr),
                           dynamicDims)[0];
 }
 
for (auto d :
        llvm::enumerate(dstStaticShape.slice(startPos, endPos - startPos + 1))) {
     if (d.index() + startPos == static_cast<unsigned>(dimIndex))
       continue;
     assert(!ShapedType::isDynamic(d.value()) &&
            "single dimension cannot be expanded into multiple dynamic "
            "dimensions");
     linearizedStaticDim *= d.value();
   }
-  Value sourceDim = builder.create<memref::DimOp>(loc, src, sourceDimPos);
+  Value sourceDim = builder.create<tensor::DimOp>(loc, src, sourceDimPos);
   return applyMapToValues(
       builder, loc,
       AffineMap::get(
           0, 1, builder.getAffineSymbolExpr(0).floorDiv(linearizedStaticDim)),
       sourceDim)[0];
 }
 
 /// Given the `src` of an expanding reshape op, the reassociation maps and the

mlir/lib/Dialect/Linalg/Transforms/ComprehensiveBufferize.cpp

if (auto bbArg = shapedValue.dyn_cast<BlockArgument>()) {
     b.setInsertionPointAfter(shapedValue.getDefiningOp());
     loc = shapedValue.getDefiningOp()->getLoc();
   }
 
   // Compute the dynamic part of the shape.
   SmallVector<Value> dynShape;
   for (auto dim : enumerate(memRefType.getShape()))
     if (dim.value() == ShapedType::kDynamicSize)
-      dynShape.push_back(
-          b.create<memref::DimOp>(loc, shapedValue, dim.index()));
+      dynShape.push_back(createOrFoldDimOp(b, loc, shapedValue, dim.index()));
 
   Value allocated = b.create<memref::AllocOp>(loc, allocMemRefType, dynShape);
   Value casted = allocated;
   if (memRefType != allocMemRefType)
     casted = b.create<memref::CastOp>(loc, memRefType, allocated);
   b.setInsertionPoint(allocated.getParentBlock()->getTerminator());
   b.create<memref::DeallocOp>(loc, allocated);
   return casted;
static LogicalResult bufferize(OpBuilder &b, LinalgOp op,
 
   // The original op will be DCE'd away later.
 
   return success();
 }
 
 /// DimOp tensor operand is modified inplace. This allows leaving dead
 /// tensors behind that will get DCE'd.
-static LogicalResult bufferize(OpBuilder &b, memref::DimOp dimOp,
+static LogicalResult bufferize(OpBuilder &b, tensor::DimOp dimOp,
                                BlockAndValueMapping &bvm,
                                const BufferizationAliasInfo &aliasInfo) {
-  if (dimOp.memrefOrTensor().getType().isa<RankedTensorType>()) {
-    Value v = lookup(bvm, dimOp.memrefOrTensor());
+  if (dimOp.source().getType().isa<RankedTensorType>()) {
+    Value v = lookup(bvm, dimOp.source());
     if (!v)
       return failure();
-    dimOp.memrefOrTensorMutable().assign(v);
+    dimOp.sourceMutable().assign(v);
   }
   return success();
 }
 
 static LogicalResult bufferize(OpBuilder &b, scf::ForOp forOp,
                                BlockAndValueMapping &bvm,
                                const BufferizationAliasInfo &aliasInfo) {
   // Take a guard before anything else.
bufferizeFuncOpInternals(FuncOp funcOp, BlockAndValueMapping &bvm,
   WalkResult result = funcOp.walk<WalkOrder::PreOrder>([&](Operation *op) {
     LogicalResult status =
         TypeSwitch<Operation *, LogicalResult>(op)
             // Skip BufferCast and TensorLoad ops.
             // clang-format off
             .Case<memref::BufferCastOp,
                   memref::TensorLoadOp>(
                 [&](auto) { return success(); })
-            .Case<memref::DimOp,
-                  scf::ForOp,
+            .Case<scf::ForOp,
+                  tensor::DimOp,
                   LinalgOp,
                   ReturnOp,
                   ExtractSliceOp,
                   InsertSliceOp,
                   VectorTransferOpInterface,
                   scf::YieldOp>(
                 [&](auto op) {
                   LDBG("Begin buferize:\n" << op << '\n');

mlir/lib/Dialect/Linalg/Transforms/Fusion.cpp

if (it != fusedLoopsAndRanges.end()) {
       ivs.push_back(it->second.offset);
       tileSizes.push_back(it->second.size);
       sizeBounds.push_back(nullptr);
       loopRanges.push_back(it->second);
       LLVM_DEBUG(llvm::dbgs() << "tiled loop#" << i << " with LoopRange "
                               << loopRanges.back() << "\n");
     } else {
       auto shapeDim = getShapeDefiningLoopRange(producer, i);
-      Value dim = b.createOrFold<memref::DimOp>(loc, shapeDim.shape,
-                                                shapeDim.dimension);
+      Value dim = createOrFoldDimOp(b, loc, shapeDim.shape, shapeDim.dimension);
       tileSizes.push_back(zero);
       sizeBounds.push_back(dim);
       loopRanges.push_back(Range{zero, dim, one});
       LLVM_DEBUG(llvm::dbgs() << "full loop#" << i << " with LoopRange "
                               << loopRanges.back() << "\n");
     }
   }
 

mlir/lib/Dialect/Linalg/Transforms/FusionOnTensors.cpp

for (OpOperand *opOperand : op.getOutputOperands()) {
         auto definingOp = operandVal.getDefiningOp<InitTensorOp>();
         if (definingOp)
           continue;
         modifiedOutput = true;
         SmallVector<Value> dynamicDims;
         for (auto dim : llvm::enumerate(operandType.getShape())) {
           if (dim.value() != ShapedType::kDynamicSize)
             continue;
-          dynamicDims.push_back(rewriter.createOrFold<memref::DimOp>(
+          dynamicDims.push_back(rewriter.createOrFold<tensor::DimOp>(
               loc, operandVal, dim.index()));
         }
         Value initTensor = rewriter.create<InitTensorOp>(
             loc, dynamicDims, operandType.getShape(),
             operandType.getElementType());
         op->setOperand(opOperand->getOperandNumber(), initTensor);
       }
     }

mlir/lib/Dialect/Linalg/Transforms/Loops.cpp

for (auto iter : llvm::enumerate(indices)) {
     }
 
     Value leftOutOfBound =
         b.create<CmpIOp>(loc, CmpIPredicate::slt, dim, zeroIndex);
     if (conds.empty())
       conds.push_back(leftOutOfBound);
     else
       conds.push_back(b.create<OrOp>(loc, conds.back(), leftOutOfBound));
-    Value rightBound = b.create<memref::DimOp>(loc, input, idx);
+    Value rightBound = createOrFoldDimOp(b, loc, input, idx);
     Value rightOutOfBound =
         b.create<CmpIOp>(loc, CmpIPredicate::sge, dim, rightBound);
     conds.push_back(b.create<OrOp>(loc, conds.back(), rightOutOfBound));
 
     // When padding is involved, the indices will only be shifted to negative,
     // so having a max op is enough.
     MLIRContext *ctx = input.getContext();
     AffineExpr m = getAffineDimExpr(/*position=*/0, ctx),

 };
 
 template <typename LoopType>
 static void lowerLinalgToLoopsImpl(FuncOp funcOp) {
   MLIRContext *context = funcOp.getContext();
   RewritePatternSet patterns(context);
   patterns.add<LinalgRewritePattern<LoopType>>(context);
   memref::DimOp::getCanonicalizationPatterns(patterns, context);
+  tensor::DimOp::getCanonicalizationPatterns(patterns, context);
   AffineApplyOp::getCanonicalizationPatterns(patterns, context);
   patterns.add<FoldAffineOp>(context);
   // Just apply the patterns greedily.
   (void)applyPatternsAndFoldGreedily(funcOp, std::move(patterns));
 }
 
 struct LowerToAffineLoops
     : public LinalgLowerToAffineLoopsBase<LowerToAffineLoops> {

mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp

static LogicalResult rewriteAsPaddedOp(PatternRewriter &rewriter,
   SmallVector<Value> paddedSubviewResults;
   paddedSubviewResults.reserve(opToPad->getNumResults());
   SetVector<Operation *> newUsersOfOpToPad;
   for (auto it : llvm::zip(opToPad->getResults(), paddedOp->getResults())) {
     auto rank = std::get<0>(it).getType().cast<RankedTensorType>().getRank();
     SmallVector<OpFoldResult> offsets(rank, rewriter.getIndexAttr(0));
     auto sizes = llvm::to_vector<4>(llvm::map_range(
         llvm::seq<unsigned>(0, rank), [&](unsigned d) -> OpFoldResult {
-          auto dimOp = rewriter.create<memref::DimOp>(loc, std::get<0>(it), d);
+          auto dimOp = rewriter.create<tensor::DimOp>(loc, std::get<0>(it), d);
           newUsersOfOpToPad.insert(dimOp);
           return dimOp.getResult();
         }));
     SmallVector<OpFoldResult> strides(rank, rewriter.getIndexAttr(1));
     paddedSubviewResults.push_back(rewriter.create<tensor::ExtractSliceOp>(
         loc, std::get<1>(it), offsets, sizes, strides));
   }
   // Replace the transient `opToPad` locally, except for uses that we just
LogicalResult ExtractSliceOfPadTensorSwapPattern::matchAndRewrite(
   // is true if the original data source turns out to be unused at runtime.
   Value dynHasZeroLenCond;
 
   int64_t rank = padOp.getSourceType().getRank();
   for (unsigned dim = 0; dim < rank; ++dim) {
     auto low = asValue(rewriter, loc, padOp.getMixedLowPad()[dim]);
     auto offset = asValue(rewriter, loc, sliceOp.getMixedOffsets()[dim]);
     auto length = asValue(rewriter, loc, sliceOp.getMixedSizes()[dim]);
-    auto srcSize = rewriter.createOrFold<memref::DimOp>(
-        loc, padOp.source(), dim);
+    auto srcSize =
+        rewriter.createOrFold<tensor::DimOp>(loc, padOp.source(), dim);
 
     // The new amount of low padding is `low - offset`. Except for the case
     // where none of the low padding is read. In that case, the new amount of
     // low padding is zero.
     Value newLow = max(zero, sub(low, offset));
     appendIndex(newLow, newLows, staticNewLows);
 
     // Start reading the data from position `offset - low`. Since the original

mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp

LogicalResult matchAndRewrite(PadTensorOp padOp,
 
     auto resultType = padOp.getResultType();
     // Compute size of InitTensorOp. Any combination of static/dynamic is
     // supported.
     SmallVector<Value> dynSizes;
     SmallVector<int64_t> staticSizes;
     for (unsigned dim = 0; dim < resultType.getRank(); ++dim) {
       if (resultType.isDynamicDim(dim)) {
-        auto srcSize = rewriter.createOrFold<memref::DimOp>(
+        auto srcSize = rewriter.createOrFold<tensor::DimOp>(
             padOp.getLoc(), padOp.source(), dim);
         // Add low and high padding value.
         auto plusLow = rewriter.createOrFold<AddIOp>(
             padOp.getLoc(), srcSize, getIdxValue(padOp.getMixedLowPad()[dim]));
         auto plusHigh = rewriter.createOrFold<AddIOp>(
             padOp.getLoc(), plusLow, getIdxValue(padOp.getMixedHighPad()[dim]));
         dynSizes.push_back(plusHigh);
       }
LogicalResult matchAndRewrite(PadTensorOp padOp,
     // PadTensorOps cannot be vectorized. Generate a InsertSliceOp instead
     // for copying the PadOp source.
 
     auto sourceType = padOp.getSourceType();
     // Compute size of source of PadTensorOp.
     SmallVector<OpFoldResult> srcSizes;
     for (unsigned dim = 0; dim < sourceType.getRank(); ++dim) {
       if (sourceType.isDynamicDim(dim)) {
-        srcSizes.push_back(rewriter.createOrFold<memref::DimOp>(
+        srcSizes.push_back(rewriter.createOrFold<tensor::DimOp>(
             padOp.getLoc(), padOp.source(), dim));
       } else {
         srcSizes.push_back(rewriter.getIndexAttr(sourceType.getDimSize(dim)));
       }
     }
     // Strides of InsertSliceOp are all 1.
     SmallVector<OpFoldResult> strides(sourceType.getRank(),
                                       rewriter.getIndexAttr(1));

mlir/lib/Dialect/Linalg/Utils/Utils.cpp

IntegerAttr getSmallestBoundingIndex(Value size) {
   } else if (auto constIndexOp = size.getDefiningOp<ConstantOp>()) {
     if (constIndexOp.getType().isa<IndexType>())
       boundingConst = constIndexOp.value().cast<IntegerAttr>().getInt();
   } else if (auto affineApplyOp = size.getDefiningOp<AffineApplyOp>()) {
     if (auto cExpr = affineApplyOp.getAffineMap()
                          .getResult(0)
                          .dyn_cast<AffineConstantExpr>())
       boundingConst = cExpr.getValue();
-  } else if (auto dimOp = size.getDefiningOp<memref::DimOp>()) {
-    auto shape = dimOp.memrefOrTensor().getType().dyn_cast<ShapedType>();
+  } else if (auto dimOp = size.getDefiningOp<tensor::DimOp>()) {
+    auto shape = dimOp.source().getType().dyn_cast<ShapedType>();
     if (auto constOp = dimOp.index().getDefiningOp<ConstantOp>()) {
       if (auto indexAttr = constOp.value().dyn_cast<IntegerAttr>()) {
         auto dimIndex = indexAttr.getInt();
         if (!shape.isDynamicDim(dimIndex)) {
           boundingConst = shape.getShape()[dimIndex];
         }
       }
     }
for (OpOperand *opOperand : linalgOp.getInputAndOutputOperands()) {
     SmallVector<OpFoldResult, 4> offsets, sizes, strides;
     offsets.reserve(rank);
     sizes.reserve(rank);
     strides.reserve(rank);
     for (unsigned r = 0; r < rank; ++r) {
       LLVM_DEBUG(llvm::dbgs() << "makeTiledShapes: for dim#" << r);
       if (!isTiled(map.getSubMap({r}), tileSizes)) {
         offsets.push_back(b.getIndexAttr(0));
-        Value dim = b.createOrFold<memref::DimOp>(loc, shapedOp, r);
+        Value dim = createOrFoldDimOp(b, loc, shapedOp, r);
         sizes.push_back(dim);
         strides.push_back(b.getIndexAttr(1));
         LLVM_DEBUG(llvm::dbgs() << ": not tiled: use size: " << dim << "\n");
         continue;
       }
       LLVM_DEBUG(llvm::dbgs() << ": tiled: figure out subsize...\n");
 
       // Tiling creates a new slice at the proper index, the slice step is 1
for (unsigned r = 0; r < rank; ++r) {
                                 << ": make sure in bound with affine.min\n");
         AffineExpr dim0, dim1, dim2;
         bindDims(context, dim0, dim1, dim2);
         // Compute min(size, dim - offset) to avoid out-of-bounds accesses.
         AffineMap minMap =
             AffineMap::inferFromExprList(
                 ArrayRef<ArrayRef<AffineExpr>>{{dim0, dim1 - dim2}})
                 .front();
-        Value d = b.create<memref::DimOp>(loc, shapedOp, r);
+        Value d = createOrFoldDimOp(b, loc, shapedOp, r);
         SmallVector<Value, 4> operands{size, d, offset};
         fullyComposeAffineMapAndOperands(&minMap, &operands);
         size = b.create<AffineMinOp>(loc, b.getIndexType(), minMap, operands);
       }
 
       sizes.push_back(size);
       LLVM_DEBUG(llvm::dbgs()
                  << "makeTiledShapes: new offset: " << offset << "\n");

mlir/lib/Dialect/MemRef/IR/MemRefDialect.cpp

 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Transforms/InliningUtils.h"
 
 using namespace mlir;
 using namespace mlir::memref;
 
 //===----------------------------------------------------------------------===//
 // MemRefDialect Dialect Interfaces
 //===----------------------------------------------------------------------===//

 };
 } // end anonymous namespace
 
 SmallVector<Value, 4> mlir::getDynOperands(Location loc, Value val,
                                            OpBuilder &b) {
   SmallVector<Value, 4> dynOperands;
   auto shapedType = val.getType().cast<ShapedType>();
   for (auto dim : llvm::enumerate(shapedType.getShape())) {
-    if (dim.value() == MemRefType::kDynamicSize)
-      dynOperands.push_back(b.create<memref::DimOp>(loc, val, dim.index()));
+    if (dim.value() == ShapedType::kDynamicSize)
+      dynOperands.push_back(createOrFoldDimOp(b, loc, val, dim.index()));
   }
   return dynOperands;
 }
 
+// Helper function that creates a memref::DimOp or tensor::DimOp depending on
+// the type of `source`.
+Value mlir::createOrFoldDimOp(OpBuilder &b, Location loc, Value source,
+                              int64_t dim) {
+  if (source.getType().isa<UnrankedMemRefType, MemRefType>())
+    return b.createOrFold<memref::DimOp>(loc, source, dim);
+  if (source.getType().isa<UnrankedTensorType, RankedTensorType>())
+    return b.createOrFold<tensor::DimOp>(loc, source, dim);
+  llvm_unreachable("Expected MemRefType or TensorType");
+}
+
 void mlir::memref::MemRefDialect::initialize() {
   addOperations<DmaStartOp, DmaWaitOp,
 #define GET_OP_LIST
 #include "mlir/Dialect/MemRef/IR/MemRefOps.cpp.inc"
                 >();
   addInterfaces<MemRefInlinerInterface>();
 }

mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp

LogicalResult DeallocOp::fold(ArrayRef<Attribute> cstOperands,
   /// dealloc(memrefcast) -> dealloc
   return foldMemRefCast(*this);
 }
 
 //===----------------------------------------------------------------------===//
 // DimOp
 //===----------------------------------------------------------------------===//
 
-void DimOp::build(OpBuilder &builder, OperationState &result, Value memref,
+void DimOp::build(OpBuilder &builder, OperationState &result, Value source,
                   int64_t index) {
   auto loc = result.location;
   Value indexValue = builder.create<ConstantIndexOp>(loc, index);
-  build(builder, result, memref, indexValue);
+  build(builder, result, source, indexValue);
 }
 
-void DimOp::build(OpBuilder &builder, OperationState &result, Value memref,
+void DimOp::build(OpBuilder &builder, OperationState &result, Value source,
                   Value index) {
   auto indexTy = builder.getIndexType();
-  build(builder, result, indexTy, memref, index);
+  build(builder, result, indexTy, source, index);
 }
 
 Optional<int64_t> DimOp::getConstantIndex() {
   if (auto constantOp = index().getDefiningOp<ConstantOp>())
     return constantOp.getValue().cast<IntegerAttr>().getInt();
   return {};
 }
 
 static LogicalResult verify(DimOp op) {
   // Assume unknown index to be in range.
   Optional<int64_t> index = op.getConstantIndex();
   if (!index.hasValue())
     return success();
 
   // Check that constant index is not knowingly out of range.
-  auto type = op.memrefOrTensor().getType();
+  auto type = op.source().getType();
   if (auto memrefType = type.dyn_cast<MemRefType>()) {
     if (index.getValue() >= memrefType.getRank())
       return op.emitOpError("index is out of range");
-  } else if (auto tensorType = type.dyn_cast<RankedTensorType>()) {
-    if (index.getValue() >= tensorType.getRank())
-      return op.emitOpError("index is out of range");
-  } else if (type.isa<UnrankedMemRefType>() || type.isa<UnrankedTensorType>()) {
+  } else if (type.isa<UnrankedMemRefType>()) {
     // Assume index to be in range.
   } else {
     llvm_unreachable("expected operand with memref type");
   }
   return success();
 }
 
 OpFoldResult DimOp::fold(ArrayRef<Attribute> operands) {
-  auto index = operands[1].dyn_cast_or_null<IntegerAttr>();
-
   // All forms of folding require a known index.
+  auto index = operands[1].dyn_cast_or_null<IntegerAttr>();
   if (!index)
     return {};
 
-  auto argTy = memrefOrTensor().getType();
-  // Fold if the shape extent along the given index is known.
-  if (auto shapedTy = argTy.dyn_cast<ShapedType>()) {
-    // Folding for unranked types (UnrankedMemRefType) is not supported.
-    if (!shapedTy.hasRank())
-      return {};
-    if (!shapedTy.isDynamicDim(index.getInt())) {
-      Builder builder(getContext());
-      return builder.getIndexAttr(shapedTy.getShape()[index.getInt()]);
-    }
-  }
+  auto memrefType = source().getType().dyn_cast<MemRefType>();
+  if (!memrefType)
 
-  Operation *definingOp = memrefOrTensor().getDefiningOp();
-
-  // dim(memref.tensor_load(memref)) -> dim(memref)
-  if (auto tensorLoadOp = dyn_cast_or_null<TensorLoadOp>(definingOp)) {
-    setOperand(0, tensorLoadOp.memref());
-    return getResult();
-  }
-
-  // Fold dim to the operand of tensor.generate.
-  if (auto fromElements = dyn_cast_or_null<tensor::GenerateOp>(definingOp)) {
-    auto resultType =
-        fromElements.getResult().getType().cast<RankedTensorType>();
-    // The case where the type encodes the size of the dimension is handled
-    // above.
-    assert(resultType.getShape()[index.getInt()] ==
-           RankedTensorType::kDynamicSize);
-
-    // Find the operand of the fromElements that corresponds to this index.
-    auto dynExtents = fromElements.dynamicExtents().begin();
-    for (auto dim : resultType.getShape().take_front(index.getInt()))
-      if (dim == RankedTensorType::kDynamicSize)
-        dynExtents++;
-
-    return Value{*dynExtents};
+  // Fold if the shape extent along the given index is known.
+  if (!memrefType.isDynamicDim(index.getInt())) {
+    Builder builder(getContext());
+    return builder.getIndexAttr(memrefType.getShape()[index.getInt()]);
   }
 
   // The size at the given index is now known to be a dynamic size.
   unsigned unsignedIndex = index.getValue().getZExtValue();
 
-  if (auto sliceOp = dyn_cast_or_null<tensor::ExtractSliceOp>(definingOp)) {
-    assert(sliceOp.isDynamicSize(unsignedIndex) &&
-           "Expected dynamic slice size");
-    return sliceOp.getDynamicSize(unsignedIndex);
-  }
-
   // Fold dim to the size argument for an `AllocOp`, `ViewOp`, or `SubViewOp`.
-  auto memrefType = argTy.dyn_cast<MemRefType>();
-  if (!memrefType)
-    return {};
+  Operation *definingOp = source().getDefiningOp();
 
   if (auto alloc = dyn_cast_or_null<AllocOp>(definingOp))
     return *(alloc.getDynamicSizes().begin() +
              memrefType.getDynamicDimIndex(unsignedIndex));
 
   if (auto alloca = dyn_cast_or_null<AllocaOp>(definingOp))
     return *(alloca.getDynamicSizes().begin() +
              memrefType.getDynamicDimIndex(unsignedIndex));

 namespace {
 /// Fold dim of a memref reshape operation to a load into the reshape's shape
 /// operand.
 struct DimOfMemRefReshape : public OpRewritePattern<DimOp> {
   using OpRewritePattern<DimOp>::OpRewritePattern;
 
   LogicalResult matchAndRewrite(DimOp dim,
                                 PatternRewriter &rewriter) const override {
-    auto reshape = dim.memrefOrTensor().getDefiningOp<ReshapeOp>();
+    auto reshape = dim.source().getDefiningOp<ReshapeOp>();
 
     if (!reshape)
       return failure();
 
     // Place the load directly after the reshape to ensure that the shape memref
     // was not mutated.
     rewriter.setInsertionPointAfter(reshape);
     Location loc = dim.getLoc();
     Value load = rewriter.create<LoadOp>(loc, reshape.shape(), dim.index());
     if (load.getType() != dim.getType())
       load = rewriter.create<IndexCastOp>(loc, dim.getType(), load);
     rewriter.replaceOp(dim, load);
     return success();
   }
 };
 
-/// Fold dim of a dim of a cast into the dim of the source of the tensor cast.
-template <typename CastOpTy>
+/// Fold dim of a cast into the dim of the source of the memref cast.
 struct DimOfCastOp : public OpRewritePattern<DimOp> {
   using OpRewritePattern<DimOp>::OpRewritePattern;
 
   LogicalResult matchAndRewrite(DimOp dimOp,
                                 PatternRewriter &rewriter) const override {
-    auto castOp = dimOp.memrefOrTensor().getDefiningOp<CastOpTy>();
+    auto castOp = dimOp.source().getDefiningOp<BufferCastOp>();
     if (!castOp)
       return failure();
     Value newSource = castOp.getOperand();
-    rewriter.replaceOpWithNewOp<DimOp>(dimOp, newSource, dimOp.index());
+    rewriter.replaceOpWithNewOp<tensor::DimOp>(dimOp, newSource, dimOp.index());
     return success();
   }
 };
 } // end anonymous namespace.
 
 void DimOp::getCanonicalizationPatterns(RewritePatternSet &results,
                                         MLIRContext *context) {
-  results.add<DimOfMemRefReshape, DimOfCastOp<BufferCastOp>,
-              DimOfCastOp<tensor::CastOp>>(context);
+  results.add<DimOfMemRefReshape, DimOfCastOp>(context);
 }
 
 // ---------------------------------------------------------------------------
 // DmaStartOp
 // ---------------------------------------------------------------------------
 
 void DmaStartOp::build(OpBuilder &builder, OperationState &result,
                        Value srcMemRef, ValueRange srcIndices, Value destMemRef,
if (auto bufferCast = memref().getDefiningOp<BufferCastOp>())
     // Approximate alias analysis by conservatively folding only when no there
     // is no interleaved operation.
     if (bufferCast->getBlock() == this->getOperation()->getBlock() &&
         bufferCast->getNextNode() == this->getOperation())
       return bufferCast.tensor();
   return {};
 }
 
+namespace {
+struct DimOfTensorLoadFolder : public OpRewritePattern<tensor::DimOp> {
+  using OpRewritePattern<tensor::DimOp>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(tensor::DimOp dimOp,
+                                PatternRewriter &rewriter) const override {
+    auto tensorLoadOp = dimOp.source().getDefiningOp<TensorLoadOp>();
+    if (!tensorLoadOp)
+      return failure();
+
+    rewriter.replaceOpWithNewOp<DimOp>(dimOp, tensorLoadOp.memref(),
+                                       dimOp.index());
+    return success();
+  }
+};
+} // namespace
+
+void TensorLoadOp::getCanonicalizationPatterns(RewritePatternSet &results,
+                                               MLIRContext *context) {
+  results.add<DimOfTensorLoadFolder>(context);
+}
+
 //===----------------------------------------------------------------------===//
 // TransposeOp
 //===----------------------------------------------------------------------===//
 
 /// Build a strided memref type by applying `permutationMap` tp `memRefType`.
 static MemRefType inferTransposeResultType(MemRefType memRefType,
                                            AffineMap permutationMap) {
   auto rank = memRefType.getRank();

mlir/lib/Dialect/MemRef/Transforms/ResolveShapedTypeResultDims.cpp

static Value getResultDimFromShapeInterface(OpBuilder &builder, OpResult result,
   if (auto attr = valueOrAttr.dyn_cast<Attribute>())
     return builder.createOrFold<ConstantIndexOp>(
         loc, attr.cast<IntegerAttr>().getInt());
   return valueOrAttr.get<Value>();
 }
 
 namespace {
 /// Fold dim of an operation that implements the InferShapedTypeOpInterface
-struct DimOfShapedTypeOpInterface : public OpRewritePattern<memref::DimOp> {
-  using OpRewritePattern<memref::DimOp>::OpRewritePattern;
+template <typename OpTy>
+struct DimOfShapedTypeOpInterface : public OpRewritePattern<OpTy> {
+  using OpRewritePattern<OpTy>::OpRewritePattern;
 
-  LogicalResult matchAndRewrite(memref::DimOp dimOp,
+  LogicalResult matchAndRewrite(OpTy dimOp,
                                 PatternRewriter &rewriter) const override {
-    OpResult dimValue = dimOp.memrefOrTensor().dyn_cast<OpResult>();
+    OpResult dimValue = dimOp.source().template dyn_cast<OpResult>();
     if (!dimValue)
       return failure();
     auto shapedTypeOp =
         dyn_cast<InferShapedTypeOpInterface>(dimValue.getOwner());
     if (!shapedTypeOp)
       return failure();
 
     Optional<int64_t> dimIndex = dimOp.getConstantIndex();

 struct ResolveShapedTypeResultDimsPass final
     : public ResolveShapedTypeResultDimsBase<ResolveShapedTypeResultDimsPass> {
   void runOnOperation() override;
 };
 } // namespace
 
 void memref::populateResolveShapedTypeResultDimsPatterns(
     RewritePatternSet &patterns) {
-  patterns.add<DimOfShapedTypeOpInterface>(patterns.getContext());
+  // TODO: Move tensor::DimOp pattern to the Tensor dialect.
+  patterns.add<DimOfShapedTypeOpInterface<memref::DimOp>,
+               DimOfShapedTypeOpInterface<tensor::DimOp>>(
+      patterns.getContext());
 }
 
 void ResolveShapedTypeResultDimsPass::runOnOperation() {
   RewritePatternSet patterns(&getContext());
   memref::populateResolveShapedTypeResultDimsPatterns(patterns);
   if (failed(applyPatternsAndFoldGreedily(getOperation()->getRegions(),
                                           std::move(patterns))))
     return signalPassFailure();
 }
 
 std::unique_ptr<Pass> memref::createResolveShapedTypeResultDimsPass() {
   return std::make_unique<ResolveShapedTypeResultDimsPass>();
 }

mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp

matchAndRewrite(ReturnOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
     rewriter.replaceOpWithNewOp<ReturnOp>(op, operands);
     return success();
   }
 };
 
 /// Sparse conversion rule for dimension accesses.
 class SparseTensorToDimSizeConverter
-    : public OpConversionPattern<memref::DimOp> {
+    : public OpConversionPattern<tensor::DimOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
   LogicalResult
-  matchAndRewrite(memref::DimOp op, ArrayRef<Value> operands,
+  matchAndRewrite(tensor::DimOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
     if (!operands[0].getType().isa<LLVM::LLVMPointerType>())
       return failure();
     Type resType = op.getType();
     StringRef name = "sparseDimSize";
     rewriter.replaceOpWithNewOp<CallOp>(
         op, resType, getFunc(op, name, resType, operands), operands);
     return success();

mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp

for (unsigned d = 0, rank = map.getNumResults(); d < rank; d++) {
         codegen.pointers[tensor][idx] =
             rewriter.create<ToPointersOp>(loc, ptrTp, t->get(), dim);
         codegen.indices[tensor][idx] =
             rewriter.create<ToIndicesOp>(loc, indTp, t->get(), dim);
       }
       // Find lower and upper bound in current dimension.
       Value up;
       if (shape[d] == MemRefType::kDynamicSize) {
-        up = rewriter.create<memref::DimOp>(loc, t->get(), d);
+        up = createOrFoldDimOp(rewriter, loc, t->get(), d);
         args.push_back(up);
       } else {
         up = rewriter.create<ConstantIndexOp>(loc, shape[d]);
       }
       codegen.sizes[idx] = codegen.highs[tensor][idx] = up;
     }
     // Perform the required bufferization. Dense inputs materialize
     // from the input tensors. Dense outputs need special handling.

mlir/lib/Dialect/StandardOps/Transforms/Bufferize.cpp

 //===----------------------------------------------------------------------===//
 
 #include "mlir/Transforms/Bufferize.h"
 #include "PassDetail.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/SCF/SCF.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/StandardOps/Transforms/Passes.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/BlockAndValueMapping.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 using namespace mlir;
 
 namespace {
-class BufferizeDimOp : public OpConversionPattern<memref::DimOp> {
+class BufferizeDimOp : public OpConversionPattern<tensor::DimOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
   LogicalResult
-  matchAndRewrite(memref::DimOp op, ArrayRef<Value> operands,
+  matchAndRewrite(tensor::DimOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
-    memref::DimOp::Adaptor adaptor(operands);
-    rewriter.replaceOpWithNewOp<memref::DimOp>(op, adaptor.memrefOrTensor(),
+    tensor::DimOp::Adaptor adaptor(operands);
+    rewriter.replaceOpWithNewOp<memref::DimOp>(op, adaptor.source(),
                                                adaptor.index());
     return success();
   }
 };
 
 class BufferizeIndexCastOp : public OpConversionPattern<IndexCastOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
void runOnFunction() override {
     // as that boils down to a select over memref descriptors (don't need to
     // touch the data).
     target.addDynamicallyLegalOp<IndexCastOp>(
         [&](IndexCastOp op) { return typeConverter.isLegal(op.getType()); });
     target.addDynamicallyLegalOp<SelectOp>([&](SelectOp op) {
       return typeConverter.isLegal(op.getType()) ||
              !op.condition().getType().isa<IntegerType>();
     });
-    target.addDynamicallyLegalOp<memref::DimOp>(
-        [&](memref::DimOp op) { return typeConverter.isLegal(op); });
     if (failed(
             applyPartialConversion(getFunction(), target, std::move(patterns))))
       signalPassFailure();
   }
 };
 } // namespace
 
 std::unique_ptr<Pass> mlir::createStdBufferizePass() {
   return std::make_unique<StdBufferizePass>();
 }

mlir/lib/Dialect/Tensor/IR/TensorOps.cpp

 #include "mlir/IR/Matchers.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/IR/TypeUtilities.h"
 #include "llvm/ADT/STLExtras.h"
 
 using namespace mlir;
 using namespace mlir::tensor;
 
+/// Materialize a single constant operation from a given attribute value with
+/// the desired resultant type.
+Operation *TensorDialect::materializeConstant(OpBuilder &builder,
+                                              Attribute value, Type type,
+                                              Location loc) {
+  return builder.create<mlir::ConstantOp>(loc, type, value);
+}
+
 //===----------------------------------------------------------------------===//
 // CastOp
 //===----------------------------------------------------------------------===//
 
 /// Determines whether tensor::CastOp casts to a more dynamic version of the
 /// source tensor. This is useful to fold a tensor.cast into a consuming op and
 /// implement canonicalization patterns for ops in different dialects that may
 /// consume the results of tensor.cast operations. Such foldable tensor.cast

 } // namespace
 
 void CastOp::getCanonicalizationPatterns(RewritePatternSet &results,
                                          MLIRContext *context) {
   results.add<ChainedTensorCast>(context);
 }
 
 //===----------------------------------------------------------------------===//
+// DimOp
+//===----------------------------------------------------------------------===//
+
+void DimOp::build(OpBuilder &builder, OperationState &result, Value source,
+                  int64_t index) {
+  auto loc = result.location;
+  Value indexValue = builder.create<ConstantIndexOp>(loc, index);
+  build(builder, result, source, indexValue);
+}
+
+void DimOp::build(OpBuilder &builder, OperationState &result, Value source,
+                  Value index) {
+  auto indexTy = builder.getIndexType();
+  build(builder, result, indexTy, source, index);
+}
+
+Optional<int64_t> DimOp::getConstantIndex() {
+  if (auto constantOp = index().getDefiningOp<ConstantOp>())
+    return constantOp.getValue().cast<IntegerAttr>().getInt();
+  return {};
+}
+
+static LogicalResult verify(DimOp op) {
+  // Assume unknown index to be in range.
+  Optional<int64_t> index = op.getConstantIndex();
+  if (!index.hasValue())
+    return success();
+
+  // Check that constant index is not knowingly out of range.
+  auto type = op.source().getType();
+  if (auto tensorType = type.dyn_cast<RankedTensorType>()) {
+    if (index.getValue() >= tensorType.getRank())
+      return op.emitOpError("index is out of range");
+  } else if (type.isa<UnrankedTensorType>()) {
+    // Assume index to be in range.
+  } else {
+    llvm_unreachable("expected operand with tensor type");
+  }
+  return success();
+}
+
+OpFoldResult DimOp::fold(ArrayRef<Attribute> operands) {
+  // All forms of folding require a known index.
+  auto index = operands[1].dyn_cast_or_null<IntegerAttr>();
+  if (!index)
+    return {};
+
+  // Folding for unranked types (UnrankedTensorType) is not supported.
+  auto tensorType = source().getType().dyn_cast<RankedTensorType>();
+  if (!tensorType)
+    return {};
+
+  // Fold if the shape extent along the given index is known.
+  if (!tensorType.isDynamicDim(index.getInt())) {
+    Builder builder(getContext());
+    return builder.getIndexAttr(tensorType.getShape()[index.getInt()]);
+  }
+
+  Operation *definingOp = source().getDefiningOp();
+
+  // Fold dim to the operand of tensor.generate.
+  if (auto fromElements = dyn_cast_or_null<tensor::GenerateOp>(definingOp)) {
+    auto resultType =
+        fromElements.getResult().getType().cast<RankedTensorType>();
+    // The case where the type encodes the size of the dimension is handled
+    // above.
+    assert(resultType.getShape()[index.getInt()] ==
+           RankedTensorType::kDynamicSize);
+
+    // Find the operand of the fromElements that corresponds to this index.
+    auto dynExtents = fromElements.dynamicExtents().begin();
+    for (auto dim : resultType.getShape().take_front(index.getInt()))
+      if (dim == RankedTensorType::kDynamicSize)
+        dynExtents++;
+
+    return Value{*dynExtents};
+  }
+
+  // The size at the given index is now known to be a dynamic size.
+  unsigned unsignedIndex = index.getValue().getZExtValue();
+
+  if (auto sliceOp = dyn_cast_or_null<tensor::ExtractSliceOp>(definingOp)) {
+    assert(sliceOp.isDynamicSize(unsignedIndex) &&
+           "Expected dynamic slice size");
+    return sliceOp.getDynamicSize(unsignedIndex);
+  }
+
+  // dim(cast) -> dim
+  if (succeeded(foldTensorCast(*this)))
+    return getResult();
+
+  return {};
+}
+
+namespace {
+/// Fold dim of a cast into the dim of the source of the tensor cast.
+struct DimOfCastOp : public OpRewritePattern<DimOp> {
+  using OpRewritePattern<DimOp>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(DimOp dimOp,
+                                PatternRewriter &rewriter) const override {
+    auto castOp = dimOp.source().getDefiningOp<CastOp>();
+    if (!castOp)
+      return failure();
+    Value newSource = castOp.getOperand();
+    rewriter.replaceOpWithNewOp<DimOp>(dimOp, newSource, dimOp.index());
+    return success();
+  }
+};
+} // end anonymous namespace.
+
+void DimOp::getCanonicalizationPatterns(RewritePatternSet &results,
+                                        MLIRContext *context) {
+  results.add<DimOfCastOp>(context);
+}
+
+//===----------------------------------------------------------------------===//
 // ExtractOp
 //===----------------------------------------------------------------------===//
 
 static LogicalResult verify(ExtractOp op) {
   // Verify the # indices match if we have a ranked type.
   if (auto tensorType = op.tensor().getType().dyn_cast<RankedTensorType>())
     if (tensorType.getRank() != static_cast<int64_t>(op.indices().size()))
       return op.emitOpError("incorrect number of indices for extract_element");

mlir/lib/Dialect/Tensor/Transforms/Bufferize.cpp

matchAndRewrite(tensor::FromElementsOp op, ArrayRef<Value> operands,
     }
     rewriter.replaceOp(op, {result});
     return success();
   }
 };
 } // namespace
 
 namespace {
+class BufferizeConstantTensor : public OpConversionPattern<ConstantOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(ConstantOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    if (!op.getType().isa<RankedTensorType>())
+      return failure();
+    auto elemType = op.getType().cast<TensorType>().getElementType();
+    auto elements = op.value().cast<DenseElementsAttr>();
+    int numberOfElements = elements.getNumElements();
+    auto resultType = MemRefType::get({numberOfElements}, elemType);
+    Value result = rewriter.create<memref::AllocOp>(op.getLoc(), resultType);
+    for (auto element : llvm::enumerate(elements.getAttributeValues())) {
+      Value index =
+          rewriter.create<ConstantIndexOp>(op.getLoc(), element.index());
+      Value val =
+          rewriter.create<ConstantOp>(op.getLoc(), element.value(), elemType);
+      rewriter.create<memref::StoreOp>(op.getLoc(), val, result, index);
+    }
+    rewriter.replaceOp(op, {result});
+    return success();
+  }
+};
+} // namespace
+
+namespace {
 class BufferizeGenerateOp : public OpConversionPattern<tensor::GenerateOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
 
   LogicalResult
   matchAndRewrite(tensor::GenerateOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const final {
     // Allocate memory.
void runOnFunction() override {
     auto *context = &getContext();
     BufferizeTypeConverter typeConverter;
     RewritePatternSet patterns(context);
     ConversionTarget target(*context);
 
     populateBufferizeMaterializationLegality(target);
 
     populateTensorBufferizePatterns(typeConverter, patterns);
+    // Cannot add BufferizeConstantTensor in populateTensorBufferizePatterns
+    // because some users of populateTensorBufferizePatterns define their own
+    // ConstantOp lowering.
+    patterns.add<BufferizeConstantTensor>(typeConverter, patterns.getContext());
+
     target.addIllegalOp<tensor::CastOp, tensor::ExtractOp,
                         tensor::FromElementsOp, tensor::GenerateOp>();
     target.addLegalDialect<memref::MemRefDialect>();
-    target.addLegalDialect<StandardOpsDialect>();
+    target.addDynamicallyLegalDialect<StandardOpsDialect>(
+        [&](Operation *op) { return typeConverter.isLegal(op); });
     target.addLegalDialect<scf::SCFDialect>();
 
     if (failed(
             applyPartialConversion(getFunction(), target, std::move(patterns))))
       signalPassFailure();
   }
 };
 } // namespace
 
 std::unique_ptr<Pass> mlir::createTensorBufferizePass() {
   return std::make_unique<TensorBufferizePass>();
 }

mlir/lib/Dialect/Tensor/Transforms/PassDetail.h

 //===- PassDetail.h - GPU Pass class details --------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef DIALECT_TENSOR_TRANSFORMS_PASSDETAIL_H_
 #define DIALECT_TENSOR_TRANSFORMS_PASSDETAIL_H_
 
 #include "mlir/Pass/Pass.h"
 
 namespace mlir {
 
+namespace memref {
+class MemRefDialect;
+} // end namespace memref
+
 namespace scf {
 class SCFDialect;
 } // end namespace scf
 
 #define GEN_PASS_CLASSES
 #include "mlir/Dialect/Tensor/Transforms/Passes.h.inc"
 
 } // end namespace mlir
 
 #endif // DIALECT_TENSOR_TRANSFORMS_PASSDETAIL_H_

mlir/lib/Dialect/Vector/VectorTransforms.cpp

xferOp.zipResultAndIndexing([&](int64_t resultIdx, int64_t indicesIdx) {
     Location loc = xferOp.getLoc();
     ImplicitLocOpBuilder lb(loc, b);
     int64_t vectorSize = xferOp.getVectorType().getDimSize(resultIdx);
     auto d0 = getAffineDimExpr(0, xferOp.getContext());
     auto vs = getAffineConstantExpr(vectorSize, xferOp.getContext());
     Value sum =
         makeComposedAffineApply(b, loc, d0 + vs, xferOp.indices()[indicesIdx]);
     Value cond = createFoldedSLE(
-        b, sum, lb.create<memref::DimOp>(xferOp.source(), indicesIdx));
+        b, sum, createOrFoldDimOp(b, loc, xferOp.source(), indicesIdx));
     if (!cond)
       return;
     // Conjunction over all dims for which we are in-bounds.
     if (inBoundsCond)
       inBoundsCond = lb.create<AndOp>(inBoundsCond, cond);
     else
       inBoundsCond = cond;
   });
assert(memrefRank == alloc.getType().cast<MemRefType>().getRank() &&
          "Expected memref rank to match the alloc rank");
   ValueRange leadingIndices =
       xferOp.indices().take_front(xferOp.getLeadingShapedRank());
   SmallVector<OpFoldResult, 4> sizes;
   sizes.append(leadingIndices.begin(), leadingIndices.end());
   auto isaWrite = isa<vector::TransferWriteOp>(xferOp);
   xferOp.zipResultAndIndexing([&](int64_t resultIdx, int64_t indicesIdx) {
     using MapList = ArrayRef<ArrayRef<AffineExpr>>;
-    Value dimMemRef = lb.create<memref::DimOp>(xferOp.source(), indicesIdx);
+    Value dimMemRef =
+        createOrFoldDimOp(b, xferOp.getLoc(), xferOp.source(), indicesIdx);
     Value dimAlloc = lb.create<memref::DimOp>(alloc, resultIdx);
     Value index = xferOp.indices()[indicesIdx];
     AffineExpr i, j, k;
     bindDims(xferOp.getContext(), i, j, k);
     SmallVector<AffineMap, 4> maps =
         AffineMap::inferFromExprList(MapList{{i - j, k}});
     // affine_min(%dimMemRef - %index, %dimAlloc)
     Value affineMin = lb.create<AffineMinOp>(
LogicalResult matchAndRewrite(ConcreteOp xferOp,
     //   (in-bounds mask):
     //   [ offset + 0 .. offset + vector_length - 1 ] < [ dim .. dim ]
     //
     // TODO: when the leaf transfer rank is k > 1, we need the last `k`
     //       dimensions here.
     unsigned vecWidth = vtp.getNumElements();
     unsigned lastIndex = llvm::size(xferOp.indices()) - 1;
     Value off = xferOp.indices()[lastIndex];
-    Value dim = rewriter.create<memref::DimOp>(loc, xferOp.source(), lastIndex);
+    Value dim = createOrFoldDimOp(rewriter, loc, xferOp.source(), lastIndex);
     Value mask = buildVectorComparison(
         rewriter, xferOp, enableIndexOptimizations, vecWidth, dim, &off);
 
     if (xferOp.mask()) {
       // Intersect the in-bounds with the mask specified as an op parameter.
       mask = rewriter.create<AndOp>(loc, mask, xferOp.mask());
     }
 

mlir/test/Conversion/ShapeToStandard/shape-to-standard.mlir

 
 // -----
 
 // Convert `rank` to `dim` of the first dimension.
 // CHECK-LABEL: @rank
 // CHECK-SAME: (%[[SHAPE:.*]]: tensor<?xindex>) -> index
 func @rank(%shape : tensor<?xindex>) -> index {
   // CHECK: %[[C0:.*]] = constant 0 : index
-  // CHECK: %[[RESULT:.*]] = memref.dim %[[SHAPE]], %[[C0]]
+  // CHECK: %[[RESULT:.*]] = tensor.dim %[[SHAPE]], %[[C0]]
   // CHECK: return %[[RESULT]] : index
   %rank = shape.rank %shape : tensor<?xindex> -> index
   return %rank : index
 }
 
 // -----
 
 // Don't lower `get_extent` if it is of type `shape.size`.

 func @rank(%shape : !shape.shape) {
   // CHECK: shape.rank
   %rank = shape.rank %shape : !shape.shape -> !shape.size
   return
 }
 
 // -----
 
-// Express `get_extent` as `memref.dim` when it relies directly on the outcome of a
+// Express `get_extent` as `tensor.dim` when it relies directly on the outcome of a
 // `shape_of` operation.
 // CHECK-LABEL: @get_extent_shape_of
 // CHECK-SAME:  (%[[ARG:.*]]: tensor<2x3xf32>, %[[IDX:.*]]: index) -> index
 func @get_extent_shape_of(%arg : tensor<2x3xf32>, %idx : index) -> index {
-  // CHECK: %[[RESULT:.*]] = memref.dim %[[ARG]], %[[IDX]] : tensor<2x3xf32>
+  // CHECK: %[[RESULT:.*]] = tensor.dim %[[ARG]], %[[IDX]] : tensor<2x3xf32>
   // CHECK: return %[[RESULT]] : index
   %shape = shape.shape_of %arg : tensor<2x3xf32> -> tensor<?xindex>
   %result = shape.get_extent %shape, %idx : tensor<?xindex>, index -> index
   return %result : index
 }
 
 // -----
 
^bb0(%index : index, %extent : index, %acc: index):
       %new_acc = muli %acc, %extent : index
       shape.yield %new_acc : index
   }
   return %num_elements : index
 }
 // CHECK-NEXT: %[[INIT:.*]] = constant 1 : index
 // CHECK-NEXT: %[[C0:.*]] = constant 0 : index
 // CHECK-NEXT: %[[C1:.*]] = constant 1 : index
-// CHECK-NEXT: %[[RANK:.*]] = memref.dim %[[SHAPE]], %[[C0]] : tensor<?xindex>
+// CHECK-NEXT: %[[RANK:.*]] = tensor.dim %[[SHAPE]], %[[C0]] : tensor<?xindex>
 // CHECK-NEXT: %[[RESULT:.*]] = scf.for %[[I:.*]] = %[[C0]] to %[[RANK]] step %[[C1]] iter_args(%[[ACC:.*]] = %[[INIT]]) -> (index)
 // CHECK-NEXT:   %[[EXTENT:.*]] = tensor.extract %[[SHAPE]][%[[I]]]
 // CHECK-NEXT:   %[[NEW_ACC:.*]] = muli %[[ACC]], %[[EXTENT]] : index
 // CHECK-NEXT:   scf.yield %[[NEW_ACC]] : index
 // CHECK-NEXT: }
 // CHECK-NEXT: return %[[RESULT]] : index
 
 // -----

 
 // Lower `shape_of` for unranked tensors.
 // CHECK-LABEL: @shape_of_unranked
 // CHECK-SAME: (%[[ARG:.*]]: tensor<*xf32>)
 func @shape_of_unranked(%arg : tensor<*xf32>) {
   // CHECK: %[[RANK:.*]] = rank %[[ARG]] : tensor<*xf32>
   // CHECK: %[[SHAPE:.*]] = tensor.generate %[[RANK]] {
   // CHECK: ^bb0(%[[I:.*]]: index):
-  // CHECK:   %[[EXTENT:.*]] = memref.dim %[[ARG]], %[[I]] : tensor<*xf32>
+  // CHECK:   %[[EXTENT:.*]] = tensor.dim %[[ARG]], %[[I]] : tensor<*xf32>
   // CHECK:   yield %[[EXTENT]] : index
   // CHECK: } : tensor<?xindex>
   %shape = shape.shape_of %arg : tensor<*xf32> -> tensor<?xindex>
   return
 }
 
 // -----
 

 
 // Lower `shape_of` for dynamically shaped tensor.
 // CHECK-LABEL: @shape_of_dyn
 // CHECK-SAME: (%[[ARG:.*]]: tensor<1x5x?xf32>)
 func @shape_of_dyn(%arg : tensor<1x5x?xf32>) {
   // CHECK-DAG: %[[C1:.*]] = constant 1 : index
   // CHECK-DAG: %[[C5:.*]] = constant 5 : index
   // CHECK-DAG: %[[C2:.*]] = constant 2 : index
-  // CHECK-DAG: %[[DYN_DIM:.*]] = memref.dim %[[ARG]], %[[C2]] : tensor<1x5x?xf32>
+  // CHECK-DAG: %[[DYN_DIM:.*]] = tensor.dim %[[ARG]], %[[C2]] : tensor<1x5x?xf32>
   // CHECK-DAG: %[[SHAPE_UNCASTED:.*]] = tensor.from_elements %[[C1]], %[[C5]], %[[DYN_DIM]] : tensor<3xindex>
   %shape = shape.shape_of %arg : tensor<1x5x?xf32> -> tensor<?xindex>
   return
 }
 
 // -----
 
 // CHECK-LABEL:  @shape_eq
 // CHECK-SAME:   (%[[A:.*]]: tensor<?xindex>, %[[B:.*]]: tensor<?xindex>) -> i1
 func @shape_eq(%a : tensor<?xindex>, %b : tensor<?xindex>) -> i1 {
   // CHECK: %[[C0:.*]] = constant 0 : index
-  // CHECK: %[[RANK_A:.*]] = memref.dim %[[A]], %[[C0]] : tensor<?xindex>
-  // CHECK: %[[RANK_B:.*]] = memref.dim %[[B]], %[[C0]] : tensor<?xindex>
+  // CHECK: %[[RANK_A:.*]] = tensor.dim %[[A]], %[[C0]] : tensor<?xindex>
+  // CHECK: %[[RANK_B:.*]] = tensor.dim %[[B]], %[[C0]] : tensor<?xindex>
   // CHECK: %[[RANK_EQ:.*]] = cmpi eq, %[[RANK_A]], %[[RANK_B]]
   // CHECK: %[[SHAPE_EQ:.*]] = scf.if %[[RANK_EQ]] -> (i1) {
   // CHECK:   %[[C1:.*]] = constant 1 : index
   // CHECK:   %[[INIT:.*]] = constant true
   // CHECK:   %[[SHAPE_EQ_INNER:.*]] = scf.for %[[I:.*]] = %[[C0]] to %[[RANK_A]] step %[[C1]] iter_args(%[[CONJ:.*]] = %[[INIT]]) -> (i1) {
   // CHECK:     %[[EXTENT_A:.*]] = tensor.extract %[[A]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_B:.*]] = tensor.extract %[[B]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_EQ:.*]] = cmpi eq, %[[EXTENT_A]], %[[EXTENT_B]]

 }
 
 // -----
 
 // CHECK-LABEL:  @shape_eq
 // CHECK-SAME:   (%[[A:.*]]: tensor<?xindex>, %[[B:.*]]: tensor<?xindex>, %[[C:.*]]: tensor<?xindex>) -> i1
 func @shape_eq(%a : tensor<?xindex>, %b : tensor<?xindex>, %c : tensor<?xindex>) -> i1 {
   // CHECK: %[[C0:.*]] = constant 0 : index
-  // CHECK: %[[RANK_A:.*]] = memref.dim %[[A]], %[[C0]] : tensor<?xindex>
-  // CHECK: %[[RANK_B:.*]] = memref.dim %[[B]], %[[C0]] : tensor<?xindex>
+  // CHECK: %[[RANK_A:.*]] = tensor.dim %[[A]], %[[C0]] : tensor<?xindex>
+  // CHECK: %[[RANK_B:.*]] = tensor.dim %[[B]], %[[C0]] : tensor<?xindex>
   // CHECK: %[[RANK_EQ:.*]] = cmpi eq, %[[RANK_A]], %[[RANK_B]]
   // CHECK: %[[SHAPE_EQ:.*]] = scf.if %[[RANK_EQ]] -> (i1) {
   // CHECK:   %[[C1:.*]] = constant 1 : index
   // CHECK:   %[[INIT:.*]] = constant true
   // CHECK:   %[[SHAPE_EQ_INNER:.*]] = scf.for %[[I:.*]] = %[[C0]] to %[[RANK_A]] step %[[C1]] iter_args(%[[CONJ:.*]] = %[[INIT]]) -> (i1) {
   // CHECK:     %[[EXTENT_A:.*]] = tensor.extract %[[A]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_B:.*]] = tensor.extract %[[B]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_EQ:.*]] = cmpi eq, %[[EXTENT_A]], %[[EXTENT_B]]
   // CHECK:     %[[CONJ_NEXT:.*]] = and %[[CONJ]], %[[EXTENT_EQ]]
   // CHECK:     scf.yield %[[CONJ_NEXT]] : i1
   // CHECK:   }
   // CHECK:   scf.yield %[[SHAPE_EQ_INNER]] : i1
   // CHECK: } else {
   // CHECK:   %[[SHAPE_EQ_INNER:.*]] = constant false
   // CHECK:   scf.yield %[[SHAPE_EQ_INNER]] : i1
   // CHECK: }
-  // CHECK: %[[RANK_C:.*]] = memref.dim %[[C]], %[[C0]] : tensor<?xindex>
+  // CHECK: %[[RANK_C:.*]] = tensor.dim %[[C]], %[[C0]] : tensor<?xindex>
   // CHECK: %[[RANK_EQ:.*]] = cmpi eq, %[[RANK_A]], %[[RANK_C]]
   // CHECK: %[[SHAPE_EQ2:.*]] = scf.if %[[RANK_EQ]] -> (i1) {
   // CHECK:   %[[C1:.*]] = constant 1 : index
   // CHECK:   %[[INIT:.*]] = constant true
   // CHECK:   %[[SHAPE_EQ_INNER:.*]] = scf.for %[[I:.*]] = %[[C0]] to %[[RANK_A]] step %[[C1]] iter_args(%[[CONJ:.*]] = %[[INIT]]) -> (i1) {
   // CHECK:     %[[EXTENT_A:.*]] = tensor.extract %[[A]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_C:.*]] = tensor.extract %[[C]][%[[I]]] : tensor<?xindex>
   // CHECK:     %[[EXTENT_EQ:.*]] = cmpi eq, %[[EXTENT_A]], %[[EXTENT_C]]
func @try_is_broadcastable (%a : tensor<2xindex>, %b : tensor<3xindex>, %c : tensor<2xindex>) -> i1 {
   return %0 : i1
 }
 // CHECK-LABEL: @try_is_broadcastable
 // CHECK-SAME:          %[[ARG0:.*]]: tensor<2xindex>,
 // CHECK-SAME:          %[[ARG1:.*]]: tensor<3xindex>,
 // CHECK-SAME:          %[[ARG2:.*]]: tensor<2xindex>)
 // CHECK:           %[[C0:.*]] = constant 0 : index
 // CHECK:           %[[C1:.*]] = constant 1 : index
-// CHECK:           %[[RANK0:.*]] = memref.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
-// CHECK:           %[[RANK1:.*]] = memref.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
-// CHECK:           %[[RANK2:.*]] = memref.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK0:.*]] = tensor.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK1:.*]] = tensor.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
+// CHECK:           %[[RANK2:.*]] = tensor.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
 // CHECK:           %[[CMP0:.*]] = cmpi ugt, %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[LARGER_DIM:.*]] = select %[[CMP0]], %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[CMP1:.*]] = cmpi ugt, %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[MAX_RANK:.*]] = select %[[CMP1]], %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[DIM_DIFF0:.*]] = subi %[[MAX_RANK]], %[[RANK0]] : index
 // CHECK:           %[[DIM_DIFF1:.*]] = subi %[[MAX_RANK]], %[[RANK1]] : index
 // CHECK:           %[[DIM_DIFF2:.*]] = subi %[[MAX_RANK]], %[[RANK2]] : index
 // CHECK:           %[[TRUE:.*]] = constant true
func @broadcast(%a : tensor<2xindex>, %b : tensor<3xindex>, %c : tensor<2xindex>) -> !shape.witness {
   return %0 : !shape.witness
 }
 // CHECK-LABEL:   func @broadcast(
 // CHECK-SAME:          %[[ARG0:.*]]: tensor<2xindex>,
 // CHECK-SAME:          %[[ARG1:.*]]: tensor<3xindex>,
 // CHECK-SAME:          %[[ARG2:.*]]: tensor<2xindex>)
 // CHECK:           %[[C0:.*]] = constant 0 : index
 // CHECK:           %[[C1:.*]] = constant 1 : index
-// CHECK:           %[[RANK0:.*]] = memref.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
-// CHECK:           %[[RANK1:.*]] = memref.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
-// CHECK:           %[[RANK2:.*]] = memref.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK0:.*]] = tensor.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK1:.*]] = tensor.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
+// CHECK:           %[[RANK2:.*]] = tensor.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
 // CHECK:           %[[CMP0:.*]] = cmpi ugt, %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[LARGER_DIM:.*]] = select %[[CMP0]], %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[CMP1:.*]] = cmpi ugt, %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[MAX_RANK:.*]] = select %[[CMP1]], %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[DIM_DIFF0:.*]] = subi %[[MAX_RANK]], %[[RANK0]] : index
 // CHECK:           %[[DIM_DIFF1:.*]] = subi %[[MAX_RANK]], %[[RANK1]] : index
 // CHECK:           %[[DIM_DIFF2:.*]] = subi %[[MAX_RANK]], %[[RANK2]] : index
 // CHECK:           %[[TRUE:.*]] = constant true

 func @broadcast_3_shapes_different_extents(%a : tensor<2xindex>,
                                            %b : tensor<3xindex>,
                                            %c : tensor<2xindex>) {
 // CHECK-LABEL:   func @broadcast_3_shapes_different_extents(
 // CHECK-SAME:          %[[ARG0:.*]]: tensor<2xindex>,
 // CHECK-SAME:          %[[ARG1:.*]]: tensor<3xindex>,
 // CHECK-SAME:          %[[ARG2:.*]]: tensor<2xindex>) {
 // CHECK:           %[[C0:.*]] = constant 0 : index
-// CHECK:           %[[RANK0:.*]] = memref.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
-// CHECK:           %[[RANK1:.*]] = memref.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
-// CHECK:           %[[RANK2:.*]] = memref.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK0:.*]] = tensor.dim %[[ARG0]], %[[C0]] : tensor<2xindex>
+// CHECK:           %[[RANK1:.*]] = tensor.dim %[[ARG1]], %[[C0]] : tensor<3xindex>
+// CHECK:           %[[RANK2:.*]] = tensor.dim %[[ARG2]], %[[C0]] : tensor<2xindex>
 // CHECK:           %[[CMP0:.*]] = cmpi ugt, %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[LARGER_DIM:.*]] = select %[[CMP0]], %[[RANK1]], %[[RANK0]] : index
 // CHECK:           %[[CMP1:.*]] = cmpi ugt, %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[MAX_RANK:.*]] = select %[[CMP1]], %[[RANK2]], %[[LARGER_DIM]] : index
 // CHECK:           %[[DIM_DIFF0:.*]] = subi %[[MAX_RANK]], %[[RANK0]] : index
 // CHECK:           %[[DIM_DIFF1:.*]] = subi %[[MAX_RANK]], %[[RANK1]] : index
 // CHECK:           %[[DIM_DIFF2:.*]] = subi %[[MAX_RANK]], %[[RANK2]] : index
 // CHECK:           %[[RESULT:.*]] = tensor.generate %[[MAX_RANK]]  {

 
 // -----
 
 // Lower `split_at`
 // CHECK-LABEL: @split_at
 // CHECK-SAME: %[[SHAPE:.*]]: tensor<?xindex>, %[[INDEX:.*]]: index
 func @split_at(%shape: tensor<?xindex>, %index: index) -> (tensor<?xindex>, tensor<?xindex>) {
   // CHECK-NEXT: %[[C0:.*]] = constant 0 : index
-  // CHECK-NEXT: %[[RANK:.*]] = memref.dim %[[SHAPE]], %[[C0]] : tensor<?xindex>
+  // CHECK-NEXT: %[[RANK:.*]] = tensor.dim %[[SHAPE]], %[[C0]] : tensor<?xindex>
   // CHECK-NEXT: %[[POSINDEX:.*]] = addi %[[INDEX]], %[[RANK]] : index
   // CHECK-NEXT: %[[ISNEG:.*]] = cmpi slt, %[[INDEX]], %[[C0]] : index
   // CHECK-NEXT: %[[SELECT:.*]] = select %[[ISNEG]], %[[POSINDEX]], %[[INDEX]] : index
   // CHECK-NEXT: %[[C1:.*]] = constant 1 : index
   // CHECK-NEXT: %[[HEAD:.*]] = tensor.extract_slice %[[SHAPE]][%[[C0]]] [%[[SELECT]]] [%[[C1]]] : tensor<?xindex> to tensor<?xindex>
   // CHECK-NEXT: %[[TAIL_SIZE:.*]] = subi %[[RANK]], %[[SELECT]] : index
   // CHECK-NEXT: %[[TAIL:.*]] = tensor.extract_slice %[[SHAPE]][%[[SELECT]]] [%[[TAIL_SIZE]]] [%[[C1]]] : tensor<?xindex> to tensor<?xindex>
   // CHECK-NEXT: return %[[HEAD]], %[[TAIL]] : tensor<?xindex>, tensor<?xindex>
   %head, %tail = "shape.split_at"(%shape, %index) : (tensor<?xindex>, index) -> (tensor<?xindex>, tensor<?xindex>)
   return %head, %tail : tensor<?xindex>, tensor<?xindex>
 }

mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir

 // -----
 
 // CHECK-LABEL: @concat
 func @concat(%arg0: tensor<5x1xf32>, %arg1: tensor<6x1xf32>) -> () {
   // CHECK: [[AXIS:%.+]] = constant 0
   // CHECK: [[STRIDE:%.+]]   = constant 1
   // CHECK: [[OFFSET:%.+]] = constant 0 : index
   // CHECK: [[IDX0:%.+]] = constant 0 : index
-  // CHECK: [[ARG0_DIM0:%.+]] = memref.dim %arg0, [[IDX0]]
+  // CHECK: [[ARG0_DIM0:%.+]] = tensor.dim %arg0, [[IDX0]]
   // CHECK: [[IDX1:%.+]] = constant 1 : index
-  // CHECK: [[ARG0_DIM1:%.+]] = memref.dim %arg0, [[IDX1]]
-  // CHECK: [[ARG1_AXIS:%.+]] = memref.dim %arg1, [[AXIS]]
+  // CHECK: [[ARG0_DIM1:%.+]] = tensor.dim %arg0, [[IDX1]]
+  // CHECK: [[ARG1_AXIS:%.+]] = tensor.dim %arg1, [[AXIS]]
   // CHECK: [[RESULT_AXIS:%.+]] = addi [[ARG0_DIM0]], [[ARG1_AXIS]]
   // CHECK: [[INIT:%.+]] = linalg.init_tensor [11, 1]
   // CHECK: [[CST:%.+]] = constant 0.0
   // CHECK: [[FILL:%.+]] = linalg.fill([[CST]], [[INIT]])
-  // CHECK: [[ARG0_DIM0:%.+]] = memref.dim %arg0, [[AXIS]]
+  // CHECK: [[ARG0_DIM0:%.+]] = tensor.dim %arg0, [[AXIS]]
   // CHECK: [[INSERT0:%.+]] = tensor.insert_slice %arg0 into [[FILL]]{{\[}}[[OFFSET]], [[OFFSET]]] {{\[}}[[ARG0_DIM0]], [[ARG0_DIM1]]] {{\[}}[[STRIDE]], [[STRIDE]]]
   // CHECK: [[NEW_OFFSET:%.+]] = addi [[OFFSET]], [[ARG0_DIM0]]
-  // CHECK: [[ARG1_DIM0:%.+]] = memref.dim %arg1, [[AXIS]]
+  // CHECK: [[ARG1_DIM0:%.+]] = tensor.dim %arg1, [[AXIS]]
   // CHECK: [[INSERT1:%.+]] = tensor.insert_slice %arg1 into [[INSERT0]]{{\[}}[[NEW_OFFSET]], [[OFFSET]]] {{\[}}[[ARG1_DIM0]], [[ARG0_DIM1]]] {{\[}}[[STRIDE]], [[STRIDE]]]
   %0 = "tosa.concat"(%arg0, %arg1) { axis = 0 : i64} : (tensor<5x1xf32>, tensor<6x1xf32>)  -> (tensor<11x1xf32>)
 
   // CHECK: [[AXIS:%.+]] = constant 1
   // CHECK: [[STRIDE:%.+]]   = constant 1
   // CHECK: [[OFFSET:%.+]] = constant 0 : index
   // CHECK: [[IDX0:%.+]] = constant 0 : index
-  // CHECK: [[ARG0_DIM0:%.+]] = memref.dim %arg0, [[IDX0]]
+  // CHECK: [[ARG0_DIM0:%.+]] = tensor.dim %arg0, [[IDX0]]
   // CHECK: [[IDX1:%.+]] = constant 1 : index
-  // CHECK: [[ARG0_DIM1:%.+]] = memref.dim %arg0, [[IDX1]]
-  // CHECK: [[ARG1_AXIS:%.+]] = memref.dim %arg0, [[AXIS]]
+  // CHECK: [[ARG0_DIM1:%.+]] = tensor.dim %arg0, [[IDX1]]
+  // CHECK: [[ARG1_AXIS:%.+]] = tensor.dim %arg0, [[AXIS]]
   // CHECK: [[RESULT_AXIS:%.+]] = addi [[ARG0_DIM1]], [[ARG1_AXIS]]
   // CHECK: [[INIT:%.+]] = linalg.init_tensor [5, 2]
   // CHECK: [[CST:%.+]] = constant 0.0
   // CHECK: [[FILL:%.+]] = linalg.fill([[CST]], [[INIT]])
-  // CHECK: [[ARG0_DIM1:%.+]] = memref.dim %arg0, [[AXIS]]
+  // CHECK: [[ARG0_DIM1:%.+]] = tensor.dim %arg0, [[AXIS]]
   // CHECK: [[INSERT0:%.+]] = tensor.insert_slice %arg0 into [[FILL]]{{\[}}[[OFFSET]], [[OFFSET]]] {{\[}}[[ARG0_DIM0]], [[ARG0_DIM1]]] {{\[}}[[STRIDE]], [[STRIDE]]]
   // CHECK: [[NEW_OFFSET:%.+]] = addi [[OFFSET]], [[ARG0_DIM1]]
-  // CHECK: [[ARG1_DIM1:%.+]] = memref.dim %arg0, [[AXIS]]
+  // CHECK: [[ARG1_DIM1:%.+]] = tensor.dim %arg0, [[AXIS]]
   // CHECK: [[INSERT1:%.+]] = tensor.insert_slice %arg0 into [[INSERT0]]{{\[}}[[OFFSET]], [[NEW_OFFSET]]] {{\[}}[[ARG0_DIM0]], [[ARG1_DIM1]]] {{\[}}[[STRIDE]], [[STRIDE]]]
   %1 = "tosa.concat"(%arg0, %arg0) { axis = 1 : i64} : (tensor<5x1xf32>, tensor<5x1xf32>)  -> (tensor<5x2xf32>)
   return
 }
 
 // -----
 
 // CHECK: #[[$MAP0:.*]] = affine_map<(d0) -> (d0)>
func @fully_connected(%arg0: tensor<5x3xf32>, %arg1: tensor<6x3xf32>, %arg2: tensor<6xf32>) -> (tensor<5x6xf32>) {
   %0 = "tosa.fully_connected"(%arg0, %arg1, %arg2) : (tensor<5x3xf32>, tensor<6x3xf32>, tensor<6xf32>)  -> (tensor<5x6xf32>)
   return %0 : tensor<5x6xf32>
 }
 
 // -----
 
 func @pad_float(%arg0 : tensor<1x2xf32>) -> (tensor<4x9xf32>) {
   %0 = constant dense<[[1, 2], [3, 4]]> : tensor<2x2xi32>
-  // CHECK: [[INDEX0:%.+]] = constant 0 : index
+  // TODO: Output contains multiple "constant 1 : index".
   // CHECK: [[INDEX1:%.+]] = constant 1 : index
-  // CHECK: [[ROW0:%.+]] = constant 0 : index
-  // CHECK: [[LOW0:%.+]] = tensor.extract %cst{{\[}}[[ROW0]], [[INDEX0]]]
-  // CHECK: [[HIGH0:%.+]] = tensor.extract %cst{{\[}}[[ROW0]], [[INDEX1]]]
-  // CHECK: [[LOW0_IDX:%.+]] = index_cast %0
-  // CHECK: [[HIGH0_IDX:%.+]] = index_cast %1
-  // CHECK: [[ROW1:%.+]] = constant 1 : index
-  // CHECK: [[LOW1:%.+]] = tensor.extract %cst{{\[}}%c1_1, %c0]
-  // CHECK: [[HIGH1:%.+]] = tensor.extract %cst{{\[}}%c1_1, %c1]
-  // CHECK: [[LOW1_IDX:%.+]] = index_cast [[LOW1]]
-  // CHECK: [[HIGH1_IDX:%.+]] = index_cast [[HIGH1]]
+  // CHECK: [[INDEX2:%.+]] = constant 2 : index
+  // CHECK: [[INDEX3:%.+]] = constant 3 : index
+  // CHECK: [[INDEX4:%.+]] = constant 4 : index
   // CHECK: [[CST:%.+]] = constant 0.000000e+00 : f32
-  // CHECK: %8 = linalg.pad_tensor %arg0 low{{\[}}[[LOW0_IDX]], [[LOW1_IDX]]] high{{\[}}[[HIGH0_IDX]], [[HIGH1_IDX]]]  {
+  // CHECK: linalg.pad_tensor %arg0 low{{\[}}%{{.*}}, [[INDEX3]]] high{{\[}}[[INDEX2]], [[INDEX4]]]  {
   // CHECK: ^bb0(%arg1: index, %arg2: index):  // no predecessors
   // CHECK:   linalg.yield [[CST]]
   // CHECK: } : tensor<1x2xf32> to tensor<4x9xf32>
   %1 = "tosa.pad"(%arg0, %0)  : (tensor<1x2xf32>, tensor<2x2xi32>)  -> (tensor<4x9xf32>)
   return %1 : tensor<4x9xf32>
 }
 
 func @pad_int(%arg0 : tensor<1x2xi32>) -> (tensor<4x9xi32>) {

mlir/test/Dialect/Linalg/bufferize.mlir

 // Check that the allocs properly consider the different shapes of the output
 // operands. The permuted indexing maps translate to different output shapes.
 
 // CHECK-LABEL:   func @dynamic_results(
 // CHECK-SAME:                          %[[ARG:.*]]: tensor<?x?xf32>
 // CHECK-DAG:       %[[C0:.*]] = constant 0 : index
 // CHECK-DAG:       %[[C1:.*]] = constant 1 : index
 // CHECK:           %[[MEMREF_ARG:.*]] = memref.buffer_cast %[[ARG]] : memref<?x?xf32>
-// CHECK:           %[[DIM0:.*]] = memref.dim %[[ARG]], %[[C0]] : tensor<?x?xf32>
-// CHECK:           %[[DIM1:.*]] = memref.dim %[[ARG]], %[[C1]] : tensor<?x?xf32>
+// CHECK:           %[[DIM0:.*]] = tensor.dim %[[ARG]], %[[C0]] : tensor<?x?xf32>
+// CHECK:           %[[DIM1:.*]] = tensor.dim %[[ARG]], %[[C1]] : tensor<?x?xf32>
 // CHECK:           %[[RESULT0:.*]] = memref.alloc(%[[DIM0]], %[[DIM1]]) : memref<?x?xf32>
 // CHECK:           %[[RESULT1:.*]] = memref.alloc(%[[DIM0]], %[[DIM1]]) : memref<?x?xf32>
 // CHECK:           linalg.generic
 // CHECK-SAME:      ins(%[[MEMREF_ARG]] : memref<?x?xf32>)
 // CHECK-SAME:      outs(%[[RESULT0]], %[[RESULT1]] : memref<?x?xf32>, memref<?x?xf32>)
 func @dynamic_results(%arg0: tensor<?x?xf32>)
          -> (tensor<?x?xf32>, tensor<?x?xf32>) {
     %0, %1 = linalg.generic {
func @bufferize_insert_slice(%t : tensor<?x?xf32>, %st0 : tensor<2x3xf32>, %st1 : tensor<2x?xf32>) ->
   // CHECK-DAG: %[[C0:.*]] = constant 0 : index
   // CHECK-DAG: %[[C1:.*]] = constant 1 : index
   %i0 = call @make_index() : () -> index
   // CHECK: %[[IDX:.*]] = call @make_index() : () -> index
 
 
   // CHECK-DAG: %[[M:.*]] = memref.buffer_cast %[[T]] : memref<?x?xf32>
   // CHECK-DAG: %[[SM0:.*]] = memref.buffer_cast %[[ST0]] : memref<2x3xf32>
-  // CHECK-NEXT: %[[DIM0:.*]] = memref.dim %[[T]], %[[C0]] : tensor<?x?xf32>
-  // CHECK-NEXT: %[[DIM1:.*]] = memref.dim %[[T]], %[[C1]] : tensor<?x?xf32>
+  // CHECK-NEXT: %[[DIM0:.*]] = tensor.dim %[[T]], %[[C0]] : tensor<?x?xf32>
+  // CHECK-NEXT: %[[DIM1:.*]] = tensor.dim %[[T]], %[[C1]] : tensor<?x?xf32>
   // CHECK-NEXT: %[[M_COPY0:.*]] = memref.alloc(%[[DIM0]], %[[DIM1]]) : memref<?x?xf32>
   // CHECK-NEXT: linalg.copy(%[[M]], %[[M_COPY0]]) : memref<?x?xf32>, memref<?x?xf32>
   // CHECK-NEXT: %[[SUBVIEW0:.*]] = memref.subview %[[M_COPY0]][0, 0] [2, 3] [1, 1]
   // CHECK-SAME:   memref<?x?xf32> to memref<2x3xf32, #[[$MAP0]]>
   // CHECK-NEXT: linalg.copy(%[[SM0]], %[[SUBVIEW0]]) : memref<2x3xf32>, memref<2x3xf32, #[[$MAP0]]>
   // CHECK-NEXT: %[[RT0:.*]] = memref.tensor_load %[[M_COPY0]] : memref<?x?xf32>
   %t0 = tensor.insert_slice %st0 into %t[0, 0][2, 3][1, 1] : tensor<2x3xf32> into tensor<?x?xf32>
 

mlir/test/Dialect/Linalg/canonicalize.mlir

%1 = linalg.tensor_expand_shape %0 [[0, 1], [2], [3, 4, 5]]
       : tensor<6x5x?xf32> into tensor<2x3x5x4x?x7xf32>
   return %1 : tensor<2x3x5x4x?x7xf32>
 }
 //      CHECK: #[[MAP:.+]] = affine_map<()[s0] -> (s0 floordiv 28)>
 //      CHECK: func @init_tensor_reshape_expansion
 // CHECK-SAME:   %[[ARG0:.+]]: index
 //      CHECK:   %[[C2:.+]] = constant 2
 //      CHECK:   %[[INIT1:.+]] = linalg.init_tensor [6, 5, %[[ARG0]]]
-//      CHECK:   %[[D0:.+]] = memref.dim %[[INIT1]], %[[C2]]
+//      CHECK:   %[[D0:.+]] = tensor.dim %[[INIT1]], %[[C2]]
 //      CHECK:   %[[T0:.+]] = affine.apply #[[MAP]]()[%[[D0]]]
 //      CHECK:   %[[INIT2:.+]] = linalg.init_tensor [2, 3, 5, 4, %[[T0]], 7]
 //      CHECK:   return %[[INIT2]]
 
 // -----
 
 func @init_tensor_reshape_collapse(%arg0 : index) -> tensor<6x5x?xf32> {
   %0 = linalg.init_tensor [2, 3, 5, 4, %arg0, 7] : tensor<2x3x5x4x?x7xf32>
   %1 = linalg.tensor_collapse_shape %0 [[0, 1], [2], [3, 4, 5]]
       : tensor<2x3x5x4x?x7xf32> into tensor<6x5x?xf32>
   return %1 : tensor<6x5x?xf32>
 }
 //      CHECK: #[[MAP:.+]] = affine_map<()[s0] -> (s0 * 28)>
 //      CHECK: func @init_tensor_reshape_collapse
 // CHECK-SAME:   %[[ARG0:.+]]: index
 //      CHECK:   %[[C4:.+]] = constant 4
 //      CHECK:   %[[INIT1:.+]] = linalg.init_tensor [2, 3, 5, 4, %[[ARG0]], 7]
-//      CHECK:   %[[D0:.+]] = memref.dim %[[INIT1]], %[[C4]]
+//      CHECK:   %[[D0:.+]] = tensor.dim %[[INIT1]], %[[C4]]
 //      CHECK:   %[[T0:.+]] = affine.apply #[[MAP]]()[%[[D0]]]
 //      CHECK:   %[[INIT2:.+]] = linalg.init_tensor [6, 5, %[[T0]]]
 //      CHECK:   return %[[INIT2]]
 
 // -----
 
 #map = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
 func @remove_no_op(%arg0 : tensor<?x?x?xf32>, %arg1 : tensor<?x?x?xf32>)
   -> (tensor<?x?x?xf32>, tensor<?x?x?xf32>) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?x?xf32>
-  %2 = memref.dim %arg0, %c2 : tensor<?x?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?x?xf32>
+  %2 = tensor.dim %arg0, %c2 : tensor<?x?x?xf32>
   %3 = linalg.init_tensor [%0, %1, %2] : tensor<?x?x?xf32>
   %4, %5 = linalg.generic {
     indexing_maps = [#map, #map, #map, #map],
     iterator_types = ["parallel", "parallel", "parallel"]
   } ins(%arg0, %arg1 : tensor<?x?x?xf32>, tensor<?x?x?xf32>)
     outs(%3, %3 : tensor<?x?x?xf32>, tensor<?x?x?xf32>) {
   ^bb0(%arg2 : f32, %arg3 : f32, %arg4 : f32, %arg5 : f32):
     linalg.yield %arg3, %arg2 : f32, f32
   } -> (tensor<?x?x?xf32>, tensor<?x?x?xf32>)
   return %4, %5 : tensor<?x?x?xf32>, tensor<?x?x?xf32>
 }
 // CHECK-LABEL: func @remove_no_op
 //  CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?x?xf32>
 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?x?x?xf32>
 //       CHECK:     return %[[ARG1]], %[[ARG0]]
 
 // -----
 
 #map = affine_map<(d0, d1) -> (d0, d1)>
 func @keep_not_noop(%arg0 : tensor<?x?xf32>) -> tensor<?x?xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %cst = constant 1.000000e+00 : f32
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   br ^bb1(%cst : f32)
 
 ^bb1(%arg1 : f32):
   %3 = linalg.generic
     {indexing_maps = [#map, #map], iterator_types = ["parallel", "parallel"]}
     ins(%arg0 : tensor<?x?xf32>) outs(%2 : tensor<?x?xf32>) {
     ^bb0(%arg2: f32, %arg3 : f32):
       linalg.yield %arg1 : f32
     } -> tensor<?x?xf32>
   return %3 : tensor<?x?xf32>
 }
 // CHECK-LABEL: func @keep_not_noop
 //       CHECK:   %[[RESULT:.+]] = linalg.generic
 //       CHECK:   return %[[RESULT]]
 
 // -----
 
 #map = affine_map<(d0, d1) -> (d0, d1)>
 func @keep_not_noop(%arg0 : tensor<?x?xf32>, %arg1 : tensor<?x?xf32>)
   -> (tensor<?x?xf32>, tensor<?x?xf32>) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %cst = constant 1.000000e+00 : f32
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   br ^bb1(%cst : f32)
 
 ^bb1(%arg2 : f32):
   %3:2 = linalg.generic
     {indexing_maps = [#map, #map, #map, #map],
      iterator_types = ["parallel", "parallel"]}
     ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)

 func @propogate_casts(%arg0 : tensor<?x?xf32>, %arg1 : f32, %arg2 : index,
     %arg3 : index) -> tensor<?x?xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c21 = constant 21 : index
   %c42 = constant 42 : index
   %0 = linalg.init_tensor [%c21, %c42] : tensor<?x?xf32>
   %1 = linalg.fill(%arg1, %0) : f32, tensor<?x?xf32> -> tensor<?x?xf32>
-  %2 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %3 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %2 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %3 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %4 = tensor.insert_slice %arg0 into %1[%arg2, %arg3] [%2, %3] [1, 1] : tensor<?x?xf32> into tensor<?x?xf32>
   return %4 : tensor<?x?xf32>
 }
 // CHECK-LABEL: func @propogate_casts
 //       CHECK:   %[[INIT:.+]] = linalg.init_tensor [21, 42]
 //       CHECK:   %[[FILL:.+]] = linalg.fill(%{{.+}}, %[[INIT]])
 //       CHECK:   %[[INSERTED:.+]] = tensor.insert_slice %{{.+}} into %[[FILL]]
 //       CHECK:   %[[RESULT:.+]] = tensor.cast %[[INSERTED]]

mlir/test/Dialect/Linalg/convert-elementwise-to-linalg.mlir

 // -----
 
 // Check proper init_tensor usage in a mixed case.
 // CHECK-LABEL: func @cmpf(
 //  CHECK-SAME:   %[[ARG0:[0-9a-zA-Z]*]]: tensor<4x?x?x8x2x?xf32>
 //  CHECK-SAME:   %[[ARG1:[0-9a-zA-Z]*]]: tensor<4x?x?x8x2x?xf32>
 func @cmpf(%arg0: tensor<4x?x?x8x2x?xf32>, %arg1: tensor<4x?x?x8x2x?xf32>) -> tensor<4x?x?x8x2x?xi1> {
   // CHECK: %[[C1:.*]] = constant 1 : index
-  // CHECK: %[[D1:.*]] = memref.dim %[[ARG0]], %[[C1]] : tensor<4x?x?x8x2x?xf32>
+  // CHECK: %[[D1:.*]] = tensor.dim %[[ARG0]], %[[C1]] : tensor<4x?x?x8x2x?xf32>
   // CHECK: %[[C2:.*]] = constant 2 : index
-  // CHECK: %[[D2:.*]] = memref.dim %[[ARG0]], %[[C2]] : tensor<4x?x?x8x2x?xf32>
+  // CHECK: %[[D2:.*]] = tensor.dim %[[ARG0]], %[[C2]] : tensor<4x?x?x8x2x?xf32>
   // CHECK: %[[C5:.*]] = constant 5 : index
-  // CHECK: %[[D5:.*]] = memref.dim %[[ARG0]], %[[C5]] : tensor<4x?x?x8x2x?xf32>
+  // CHECK: %[[D5:.*]] = tensor.dim %[[ARG0]], %[[C5]] : tensor<4x?x?x8x2x?xf32>
   // CHECK: %[[INIT:.*]] = linalg.init_tensor [4, %[[D1]], %[[D2]], 8, 2, %[[D5]]] : tensor<4x?x?x8x2x?xi1>
   // CHECK: linalg.generic
   // CHECK-SAME:  ins(%[[ARG0]], %[[ARG1]]
   // CHECK-SAME: outs(%[[INIT]]
   // CHECK: ^bb0(%{{.*}}: f32, %{{.*}}: f32, %{{.*}}: i1):
   // CHECK: cmpf olt, %{{.*}}, %{{.*}} : f32
   %0 = cmpf olt, %arg0, %arg1 : tensor<4x?x?x8x2x?xf32>
   return %0 : tensor<4x?x?x8x2x?xi1>
 }
 

mlir/test/Dialect/Linalg/drop-unit-extent-dims.mlir

 // CHECK-SAME:       [0, 1], [2], [3, 4, 5, 6]
 //      CHECK:   return %[[RESULT1]], %[[RESULT2]]
 
 // -----
 
 func @unit_dim_for_reduction(%arg0: tensor<1x?x1x?xf32>) -> tensor<1x?xf32> {
   %cst = constant 1.000000e+00 : f32
   %c3 = constant 3 : index
-  %0 = memref.dim %arg0, %c3 : tensor<1x?x1x?xf32>
+  %0 = tensor.dim %arg0, %c3 : tensor<1x?x1x?xf32>
   %1 = linalg.init_tensor [1, %0] : tensor<1x?xf32>
   %2 = linalg.fill(%cst, %1) : f32, tensor<1x?xf32> -> tensor<1x?xf32>
   %3 = linalg.generic {
     indexing_maps = [affine_map<(d0, d1, d2, d3) -> (d0, d1, d2, d3)>,
                      affine_map<(d0, d1, d2, d3) -> (d0, d1)>],
     iterator_types = ["parallel", "parallel", "reduction", "reduction"]}
     ins(%arg0 : tensor<1x?x1x?xf32>)
     outs(%2 : tensor<1x?xf32>) {

 //      CHECK:   %[[RESULT_RESHAPE:.+]] = linalg.tensor_expand_shape %[[RESULT]] {{\[}}[0, 1]]
 //      CHECK:   return %[[RESULT_RESHAPE]]
 
 // -----
 
 func @unit_dim_for_reduction_inner(%arg0: tensor<?x1x?x1xf32>) -> tensor<?x1xf32> {
   %cst = constant 1.000000e+00 : f32
   %c2 = constant 2 : index
-  %0 = memref.dim %arg0, %c2 : tensor<?x1x?x1xf32>
+  %0 = tensor.dim %arg0, %c2 : tensor<?x1x?x1xf32>
   %1 = linalg.init_tensor [%0, 1] : tensor<?x1xf32>
   %2 = linalg.fill(%cst, %1) : f32, tensor<?x1xf32> -> tensor<?x1xf32>
   %3 = linalg.generic {
     indexing_maps = [affine_map<(d0, d1, d2, d3) -> (d0, d1, d2, d3)>,
                      affine_map<(d0, d1, d2, d3) -> (d0, d1)>],
     iterator_types = ["parallel", "parallel", "reduction", "reduction"]}
     ins(%arg0 : tensor<?x1x?x1xf32>)
     outs(%2 : tensor<?x1xf32>) {

mlir/test/Dialect/Linalg/fusion-elementwise-options.mlir

#ternary2Dpointwise = {
   iterator_types = ["parallel", "parallel"]
 }
 func @test_fusion_limit(
     %arg0 : tensor<?x?xf32>, %arg1 : tensor<?x?xf32>, %arg2 : tensor<?x?xf32>,
     %arg3 : tensor<?x?xf32>, %arg4 : tensor<?x?xf32>, %arg5 : tensor<?x?xf32>)
     -> tensor<?x?xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %d0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %d1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %d0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %d1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %init = linalg.init_tensor [%d0, %d1] : tensor<?x?xf32>
   %0 = linalg.generic #binary2Dpointwise
       ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%init : tensor<?x?xf32>) {
     ^bb0(%arg6 : f32, %arg7 : f32, %arg8 : f32):
        %1 = mulf %arg6, %arg7 : f32
        linalg.yield %1 : f32
     } -> tensor<?x?xf32>

mlir/test/Dialect/Linalg/fusion-sequence.mlir

 module {
   func @tensor_op_fusion(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32>,
                          %arg2: tensor<?x?xf32>, %arg3: tensor<?xf32>)
     -> tensor<?x?xf32> {
     %c0 = constant 0 : index
     %c1 = constant 1 : index
     %0 = linalg.matmul ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
-    %1 = memref.dim %0, %c0 : tensor<?x?xf32>
-    %2 = memref.dim %0, %c1 : tensor<?x?xf32>
+    %1 = tensor.dim %0, %c0 : tensor<?x?xf32>
+    %2 = tensor.dim %0, %c1 : tensor<?x?xf32>
     %3 = linalg.init_tensor [%1, %2] : tensor<?x?xf32>
     %4 = linalg.generic
       {indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>,
                         affine_map<(d0, d1) -> (d0)>,
                         affine_map<(d0, d1) -> (d0, d1)>],
        iterator_types = ["parallel", "parallel"]}
       ins(%0, %arg3 : tensor<?x?xf32>, tensor<?xf32>)
       outs(%3 : tensor<?x?xf32>) {

 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG3:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG4:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG5:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG6:[a-zA-Z0-9_]+]]: tensor<?x?xf32>) -> tensor<?x?xf32> {
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C1:.+]] = constant 1 : index
-//       CHECK:   %[[M:.+]] = memref.dim %[[ARG0]], %c0 : tensor<?x?xf32>
+//       CHECK:   %[[M:.+]] = tensor.dim %[[ARG0]], %c0 : tensor<?x?xf32>
 //       CHECK:   %[[R0:.+]] = scf.for %[[IV0:[a-zA-Z0-9_]+]] =
 //  CHECK-SAME:     iter_args(%[[ARG8:.+]] = %[[ARG6]]) -> (tensor<?x?xf32>) {
-//       CHECK:     %[[M_1:.+]] = memref.dim %[[ARG8]], %[[C0]]
+//       CHECK:     %[[M_1:.+]] = tensor.dim %[[ARG8]], %[[C0]]
 //       CHECK:     %[[TILE_M_1:.+]] = affine.min #[[MAP0]](%[[M_1]], %[[IV0]])
-//       CHECK:     %[[N3:.+]] = memref.dim %[[ARG8]], %[[C1]]
+//       CHECK:     %[[N3:.+]] = tensor.dim %[[ARG8]], %[[C1]]
 //       CHECK:     %[[STARG6:.+]] = tensor.extract_slice %[[ARG8]][%[[IV0]], 0]
 //  CHECK-SAME:       [%[[TILE_M_1]], %[[N3]]]
-//       CHECK:     %[[M_2:.+]] = memref.dim %[[ARG4]], %[[C0]]
+//       CHECK:     %[[M_2:.+]] = tensor.dim %[[ARG4]], %[[C0]]
 //       CHECK:     %[[TILE_M_2:.+]] = affine.min #[[MAP1]](%[[IV0]])[%[[M_2]], %[[M]]]
-//       CHECK:     %[[N2:.+]] = memref.dim %[[ARG4]], %[[C1]]
+//       CHECK:     %[[N2:.+]] = tensor.dim %[[ARG4]], %[[C1]]
 //       CHECK:     %[[STARG4:.+]] = tensor.extract_slice %[[ARG4]][%[[IV0]], 0]
 //  CHECK-SAME:       [%[[TILE_M_2]], %[[N2]]]
 //       CHECK:     %[[TILE_M_3:.+]] = affine.min #[[MAP1]](%[[IV0]])[%[[M]], %[[M]]]
-//       CHECK:     %[[N0:.+]] = memref.dim %[[ARG0]], %[[C1]]
+//       CHECK:     %[[N0:.+]] = tensor.dim %[[ARG0]], %[[C1]]
 //       CHECK:     %[[STARG0:.+]] = tensor.extract_slice %[[ARG0]][%[[IV0]], 0]
 //  CHECK-SAME:       [%[[TILE_M_3]], %[[N0]]]
-//       CHECK:     %[[M_3:.+]] = memref.dim %[[ARG2]], %[[C0]]
+//       CHECK:     %[[M_3:.+]] = tensor.dim %[[ARG2]], %[[C0]]
 //       CHECK:     %[[TILE_M_4:.+]] = affine.min #[[MAP1]](%[[IV0]])[%[[M_3]], %[[M]]]
-//       CHECK:     %[[N1:.+]] = memref.dim %[[ARG2]], %[[C1]]
+//       CHECK:     %[[N1:.+]] = tensor.dim %[[ARG2]], %[[C1]]
 //       CHECK:     %[[STARG2:.+]] = tensor.extract_slice %[[ARG2]][%[[IV0]], 0]
 //  CHECK-SAME:       [%[[TILE_M_4]], %[[N1]]]
 //       CHECK:     %[[T0:.+]] = linalg.matmul
 //  CHECK-SAME:       ins(%[[STARG0]], %[[ARG1]] : tensor<?x?xf32>, tensor<?x?xf32>
 //  CHECK-SAME:       ) outs(%[[STARG2]] : tensor<?x?xf32>)
 //       CHECK:     %[[T1:.+]] = linalg.matmul
 //  CHECK-SAME:       ins(%[[T0]], %arg3 : tensor<?x?xf32>, tensor<?x?xf32>
 //  CHECK-SAME:       ) outs(%[[STARG4]] : tensor<?x?xf32>)
 //       CHECK:     %[[T2:.+]] = linalg.matmul
 //  CHECK-SAME:       ins(%[[T1]], %arg5 : tensor<?x?xf32>, tensor<?x?xf32>
 //  CHECK-SAME:       ) outs(%[[STARG6]] : tensor<?x?xf32>)
 //       CHECK:     %[[R1:.+]] = tensor.insert_slice %[[T2]]
 //  CHECK-SAME:       into %[[ARG8]][%[[IV0]], 0] [%[[TILE_M_1]], %[[N3]]]
 //       CHECK:     scf.yield %[[R1]] : tensor<?x?xf32>
 //       CHECK:   }

mlir/test/Dialect/Linalg/fusion-tensor-pattern.mlir

 // CHECK-SAME:   %[[ARG3:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 // CHECK-SAME:   %[[ARG4:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 
 //  CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //  CHECK-DAG:   %[[C1:.+]] = constant 1 : index
 //  CHECK-DAG:   %[[C32:.+]] = constant 32 : index
 //  CHECK-DAG:   %[[C64:.+]] = constant 64 : index
 //  CHECK-DAG:   %[[C16:.+]] = constant 16 : index
-//  CHECK-DAG:   %[[M:.+]] = memref.dim %[[ARG0]], %[[C0]]
+//  CHECK-DAG:   %[[M:.+]] = tensor.dim %[[ARG0]], %[[C0]]
 //      CHECK:   %[[RESULT:.+]] = scf.for %[[IV0:[a-zA-Z0-9]+]] =
 // CHECK-SAME:     %[[C0]] to %[[M]] step %[[C32]]
 // CHECK-SAME:     iter_args(%[[ARG6:.+]] = %[[ARG4]]) -> (tensor<?x?xf32>) {
-//      CHECK:     %[[M_2:.+]] = memref.dim %[[ARG6]], %[[C0]]
+//      CHECK:     %[[M_2:.+]] = tensor.dim %[[ARG6]], %[[C0]]
 //      CHECK:     %[[TILE_M_2:.+]] = affine.min #[[MAP1]](%[[M_2]], %[[IV0]])
-//      CHECK:     %[[N3:.+]] = memref.dim %[[ARG6]], %[[C1]]
+//      CHECK:     %[[N3:.+]] = tensor.dim %[[ARG6]], %[[C1]]
 //      CHECK:     %[[ST_ARG6:.+]] = tensor.extract_slice %[[ARG6]][%[[IV0]], 0]
 // CHECK-SAME:       [%[[TILE_M_2]], %[[N3]]]
 //      CHECK:     %[[TILE_M_3:.+]] = affine.min #[[MAP5]](%[[IV0]])[%[[M]], %[[M]]]
-//      CHECK:     %[[N1:.+]] = memref.dim %[[ARG0]], %[[C1]]
+//      CHECK:     %[[N1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
 //      CHECK:     %[[ST_ARG0:.+]] = tensor.extract_slice %[[ARG0]][%[[IV0]], 0]
 // CHECK-SAME:       [%[[TILE_M_3]], %[[N1]]]
-//      CHECK:     %[[M_3:.+]] = memref.dim %[[ARG2]], %[[C0]]
+//      CHECK:     %[[M_3:.+]] = tensor.dim %[[ARG2]], %[[C0]]
 //      CHECK:     %[[TILE_M_4:.+]] = affine.min #[[MAP5]](%[[IV0]])[%[[M_3]], %[[M]]]
-//      CHECK:     %[[N2_2:.+]] = memref.dim %[[ARG2]], %[[C1]]
+//      CHECK:     %[[N2_2:.+]] = tensor.dim %[[ARG2]], %[[C1]]
 //      CHECK:     %[[ST_ARG2:.+]] = tensor.extract_slice %[[ARG2]][%[[IV0]], 0]
 // CHECK-SAME:       [%[[TILE_M_4]], %[[N2_2]]]
 //      CHECK:     %[[LHS:.+]] = linalg.matmul
 // CHECK-SAME:       __internal_linalg_transform__ = "after_lhs_fusion_producer"
 // CHECK-SAME:       ins(%[[ST_ARG0]], %[[ARG1]] : tensor<?x?xf32>, tensor<?x?xf32>)
 // CHECK-SAME:       outs(%[[ST_ARG2]] : tensor<?x?xf32>)
-//      CHECK:     %[[N2:.+]] = memref.dim %[[ARG1]], %[[C1]]
-//      CHECK:     %[[N3_2:.+]] = memref.dim %[[ARG3]], %[[C1]]
+//      CHECK:     %[[N2:.+]] = tensor.dim %[[ARG1]], %[[C1]]
+//      CHECK:     %[[N3_2:.+]] = tensor.dim %[[ARG3]], %[[C1]]
 //      CHECK:     %[[YIELD0:.+]] = scf.for %[[IV1:[a-zA-Z0-9]+]] =
 // CHECK-SAME:       %[[C0]] to %[[N3_2]] step %[[C64]]
 // CHECK-SAME:       iter_args(%[[ARG8:.+]] = %[[ST_ARG6]]) -> (tensor<?x?xf32>) {
 //      CHECK:       %[[YIELD1:.+]] = scf.for %[[IV2:[a-zA-Z0-9]+]] =
 // CHECK-SAME:         %[[C0]] to %[[N2]] step %[[C16]]
 // CHECK-SAME:         iter_args(%[[ARG10:.+]] = %[[ARG8]]) -> (tensor<?x?xf32>) {
 //      CHECK:         %[[TILE_N2:.+]] = affine.min #[[MAP2]](%[[IV2]])[%[[N2]]]
 //      CHECK:         %[[ST_LHS:.+]] = tensor.extract_slice %[[LHS]][0, %[[IV2]]]
 // CHECK-SAME:           [%[[TILE_M_3]], %[[TILE_N2]]]
-//      CHECK:         %[[N2_3:.+]] = memref.dim %[[ARG3]], %[[C0]]
+//      CHECK:         %[[N2_3:.+]] = tensor.dim %[[ARG3]], %[[C0]]
 //      CHECK:         %[[TILE_N2_2:.+]] = affine.min #[[MAP2]](%[[IV2]])[%[[N2_3]]]
 //      CHECK:         %[[TILE_N3:.+]] = affine.min #[[MAP3]](%[[IV1]])[%[[N3_2]]]
 //      CHECK:         %[[ST_ARG3:.+]] = tensor.extract_slice %[[ARG3]][%[[IV2]], %[[IV1]]]
 // CHECK-SAME:           [%[[TILE_N2_2]], %[[TILE_N3]]]
-//      CHECK:         %[[M_4:.+]] = memref.dim %[[ARG10]], %[[C0]]
-//      CHECK:         %[[N3_3:.+]] = memref.dim %[[ARG10]], %[[C1]]
+//      CHECK:         %[[M_4:.+]] = tensor.dim %[[ARG10]], %[[C0]]
+//      CHECK:         %[[N3_3:.+]] = tensor.dim %[[ARG10]], %[[C1]]
 //      CHECK:         %[[TILE_N3_2:.+]] = affine.min #[[MAP4]](%[[N3_3]], %[[IV1]])
 //      CHECK:         %[[ST_ARG4:.+]] = tensor.extract_slice %[[ARG10]][0, %[[IV1]]]
 // CHECK-SAME:           [%[[M_4]], %[[TILE_N3_2]]]
 //      CHECK:         %[[ST_RESULT:.+]] = linalg.matmul
 // CHECK-SAME:           __internal_linalg_transform__ = "after_lhs_fusion"
 // CHECK-SAME:           ins(%[[ST_LHS]], %[[ST_ARG3]]
 // CHECK-SAME:             : tensor<?x?xf32>, tensor<?x?xf32>)
 // CHECK-SAME:           outs(%[[ST_ARG4]] : tensor<?x?xf32>)

 // TLOOP-SAME: %[[ABC_INIT:[a-zA-Z0-9_]+]]: tensor<?x?xf32>) -> tensor<?x?xf32> {
 
 // TLOOP-DAG:  %[[C32:.*]] = constant 32 : index
 // TLOOP-DAG:  %[[C64:.*]] = constant 64 : index
 // TLOOP-DAG:  %[[C16:.*]] = constant 16 : index
 // TLOOP-DAG:  %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG:  %[[C1:.*]] = constant 1 : index
 
-// TLOOP:  %[[DIM_A0:.*]] = memref.dim %[[A]], %[[C0]] : [[TY:.*]]
+// TLOOP:  %[[DIM_A0:.*]] = tensor.dim %[[A]], %[[C0]] : [[TY:.*]]
 
 // TLOOP:  %[[ABC:.*]] = linalg.tiled_loop (%[[IV0:.*]]) = (%[[C0]]) 
 // TLOOP-SAME: to (%[[DIM_A0]]) step (%[[C32]]) 
 // TLOOP-SAME: ins (%[[C_:.*]] = %[[C]]: tensor<?x?xf32>,
 // TLOOP-SAME:      %[[A_:.*]] = %[[A]]: tensor<?x?xf32>,
 // TLOOP-SAME:      %[[B_:.*]] = %[[B]]: tensor<?x?xf32>,
 // TLOOP-SAME:      %[[AB_INIT_:.*]] = %[[AB_INIT]]: tensor<?x?xf32>)
 // TLOOP-SAME: outs (%[[ABC_INIT_:.*]] = %[[ABC_INIT]]: tensor<?x?xf32>) {
 
 // TLOOP:    %[[ABC_INIT_SUB:.*]] = tensor.extract_slice %[[ABC_INIT_]][%[[IV0]], 0]
 // TLOOP:    %[[A_SUB:.*]] = tensor.extract_slice %[[A_]][%[[IV0]], 0]
 // TLOOP:    %[[AB_INIT_SUB:.*]] = tensor.extract_slice %[[AB_INIT_]][%[[IV0]], 0]
 
 // TLOOP:    %[[AB_SUB:.*]] = linalg.matmul
 // TLOOP-SAME:  ins(%[[A_SUB]], %[[B_]] : {{.*}}) outs(%[[AB_INIT_SUB]]
 
-// TLOOP:    %[[DIM_B_1:.*]] = memref.dim %[[B_]], %[[C1]] : [[TY]]
-// TLOOP:    %[[DIM_C_1:.*]] = memref.dim %[[C_]], %[[C1]] : [[TY]]
+// TLOOP:    %[[DIM_B_1:.*]] = tensor.dim %[[B_]], %[[C1]] : [[TY]]
+// TLOOP:    %[[DIM_C_1:.*]] = tensor.dim %[[C_]], %[[C1]] : [[TY]]
 
 // TLOOP:    %[[ABC_SUB_:.*]] = linalg.tiled_loop (%[[IV1:.*]], %[[IV2:.*]]) = 
 // TLOOP-SAME: (%[[C0]], %[[C0]]) to (%[[DIM_C_1]], %[[DIM_B_1]])
 // TLOOP-SAME: step (%[[C64]], %[[C16]]) 
 // TLOOP-SAME: ins (%[[AB_SUB_:.*]] = %[[AB_SUB]]: [[TY]],
 // TLOOP-SAME:      %[[C__:.*]] = %[[C_]]: [[TY]])
 // TLOOP-SAME: outs (%[[ABC_INIT_SUB_:.*]] = %[[ABC_INIT_SUB]]: [[TY]])
 // TLOOP-SAME: iterators["parallel", "reduction"] {

 
 // -----
 
 module {
   func @matmul_plus_matmul(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32>,
                            %arg2: tensor<?x?xf32>) -> tensor<?x?xf32>{
     %c0 = constant 0 : index
     %c1 = constant 1 : index
-    %0 = memref.dim %arg2, %c0 : tensor<?x?xf32>
-    %1 = memref.dim %arg2, %c1 : tensor<?x?xf32>
+    %0 = tensor.dim %arg2, %c0 : tensor<?x?xf32>
+    %1 = tensor.dim %arg2, %c1 : tensor<?x?xf32>
     %2 = linalg.matmul ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
-    %3 = memref.dim %2, %c0 : tensor<?x?xf32>
-    %4 = memref.dim %2, %c1 : tensor<?x?xf32>
+    %3 = tensor.dim %2, %c0 : tensor<?x?xf32>
+    %4 = tensor.dim %2, %c1 : tensor<?x?xf32>
     %5 = linalg.init_tensor [%3, %4] : tensor<?x?xf32>
     %6 = linalg.generic
       {indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>,
                         affine_map<(d0, d1) -> (d0, d1)>,
                         affine_map<(d0, d1) -> (d0, d1)>],
        iterator_types = ["parallel", "parallel"],
        __internal_linalg_transform__ = "transpose_fusion"}
       ins(%2, %2 : tensor<?x?xf32>, tensor<?x?xf32>)

 // TLOOP-SAME:    %[[B:[a-zA-Z0-9_]+]]: tensor<?x?xf32>,
 // TLOOP-SAME:    %[[AB:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 
 // TLOOP-DAG:  %[[C32:.*]] = constant 32 : index
 // TLOOP-DAG:  %[[C64:.*]] = constant 64 : index
 // TLOOP-DAG:  %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG:  %[[C1:.*]] = constant 1 : index
 
-// TLOOP:  %[[DIM_A_0:.*]] = memref.dim %[[A]], %[[C0]] : [[TY:.*]]
-// TLOOP:  %[[DIM_B_1:.*]] = memref.dim %[[B]], %[[C1]] : [[TY]]
+// TLOOP:  %[[DIM_A_0:.*]] = tensor.dim %[[A]], %[[C0]] : [[TY:.*]]
+// TLOOP:  %[[DIM_B_1:.*]] = tensor.dim %[[B]], %[[C1]] : [[TY]]
 
 // TLOOP:  %[[INIT:.*]] = linalg.init_tensor [%[[DIM_A_0]], %[[DIM_B_1]]]
 
 // TLOOP:  %[[RESULT:.*]] = linalg.tiled_loop (%[[IV0:.*]], %[[IV1:.*]]) =
 // TLOOP-SAME: (%[[C0]], %[[C0]]) to (%[[DIM_A_0]], %[[DIM_B_1]])
 // TLOOP-SAME: step (%[[C32]], %[[C64]])
 // TLOOP-SAME: ins (%[[A_:.*]] = %[[A]]: [[TY]],
 // TLOOP-SAME:      %[[B_:.*]] = %[[B]]: [[TY]],

 
 // TLOOP-DAG:  %[[C0_F32:.*]] = constant 0.0
 // TLOOP-DAG:  %[[C32:.*]] = constant 32 : index
 // TLOOP-DAG:  %[[C64:.*]] = constant 64 : index
 // TLOOP-DAG:  %[[C16:.*]] = constant 16 : index
 // TLOOP-DAG:  %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG:  %[[C1:.*]] = constant 1 : index
 
-// TLOOP:  %[[DIM_A_0:.*]] = memref.dim %[[A]], %[[C0]] : [[TY:.*]]
-// TLOOP:  %[[DIM_B_1:.*]] = memref.dim %[[B]], %[[C1]] : [[TY]]
+// TLOOP:  %[[DIM_A_0:.*]] = tensor.dim %[[A]], %[[C0]] : [[TY:.*]]
+// TLOOP:  %[[DIM_B_1:.*]] = tensor.dim %[[B]], %[[C1]] : [[TY]]
 
 // TLOOP:  %[[AB:.*]] = linalg.tiled_loop (%[[I:.*]], %[[J:.*]]) =
 // TLOOP-SAME: (%[[C0]], %[[C0]]) to (%[[DIM_A_0]], %[[DIM_B_1]])
 // TLOOP-SAME: step (%[[C32]], %[[C64]])
 // TLOOP-SAME: ins (%[[A_:.*]] = %[[A]]: [[TY]],
 // TLOOP-SAME:      %[[B_:.*]] = %[[B]]: [[TY]],
 // TLOOP-SAME:      %[[C0_F32_:.*]] = %[[C0_F32]]
 // TLOOP-SAME: outs (%[[OUT_:.*]] = %[[OUT]]: [[TY]]) {
 
-// TLOOP:    %[[DIM_A__1:.*]] = memref.dim %[[A_]], %[[C1]] : [[TY]]
+// TLOOP:    %[[DIM_A__1:.*]] = tensor.dim %[[A_]], %[[C1]] : [[TY]]
 // TLOOP:    %[[A_SUB:.*]] = tensor.extract_slice %[[A_]][%[[I]], 0]
 // TLOOP:    %[[B_SUB:.*]] = tensor.extract_slice %[[B_]][0, %[[J]]]
 // TLOOP:    %[[OUT_SUB:.*]] = tensor.extract_slice %[[OUT_]][%[[I]], %[[J]]]
 // TLOOP:    %[[INIT_SUB:.*]] = linalg.fill(%[[C0_F32_]], %[[OUT_SUB]])
 
 // TLOOP:    %[[AB_SUB:.*]] = linalg.tiled_loop (%[[K:.*]]) = (%[[C0]])
 // TLOOP-SAME: to (%[[DIM_A__1]]) step (%[[C16]])
 // TLOOP-SAME: ins (%[[A_SUB_:.*]] = %[[A_SUB]]: [[TY]],

 
 // TLOOP-DAG:  %[[C0_F32:.*]] = constant 0.0
 // TLOOP-DAG:  %[[C32:.*]] = constant 32 : index
 // TLOOP-DAG:  %[[C64:.*]] = constant 64 : index
 // TLOOP-DAG:  %[[C16:.*]] = constant 16 : index
 // TLOOP-DAG:  %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG:  %[[C1:.*]] = constant 1 : index
 
-// TLOOP:  %[[DIM_A_0:.*]] = memref.dim %[[A]], %[[C0]] : [[TY:.*]]
-// TLOOP:  %[[DIM_B_1:.*]] = memref.dim %[[B]], %[[C1]] : [[TY]]
+// TLOOP:  %[[DIM_A_0:.*]] = tensor.dim %[[A]], %[[C0]] : [[TY:.*]]
+// TLOOP:  %[[DIM_B_1:.*]] = tensor.dim %[[B]], %[[C1]] : [[TY]]
 
 // TLOOP:  %[[AB:.*]] = linalg.tiled_loop (%[[I:.*]], %[[J:.*]]) =
 // TLOOP-SAME: (%[[C0]], %[[C0]]) to (%[[DIM_A_0]], %[[DIM_B_1]])
 // TLOOP-SAME: step (%[[C32]], %[[C64]])
 // TLOOP-SAME: ins (%[[A_:.*]] = %[[A]]: [[TY]],
 // TLOOP-SAME:      %[[B_:.*]] = %[[B]]: [[TY]],
 // TLOOP-SAME:      %[[C0_F32_:.*]] = %[[C0_F32]]
 // TLOOP-SAME: outs (%[[OUT_:.*]] = %[[OUT]]: [[TY]]) {
 
-// TLOOP:    %[[DIM_A__1:.*]] = memref.dim %[[A_]], %[[C1]] : [[TY]]
+// TLOOP:    %[[DIM_A__1:.*]] = tensor.dim %[[A_]], %[[C1]] : [[TY]]
 // TLOOP:    %[[A_SUB:.*]] = tensor.extract_slice %[[A_]][%[[I]], 0]
 // TLOOP:    %[[B_SUB:.*]] = tensor.extract_slice %[[B_]][0, %[[J]]]
 // TLOOP:    %[[OUT_SUB:.*]] = tensor.extract_slice %[[OUT_]][%[[I]], %[[J]]]
 // TLOOP:    %[[INIT_SUB:.*]] = linalg.generic
 // TLOOP-SAME: ins(%[[C0_F32_]]
 // TLOOP-SAME: outs(%[[OUT_SUB]]
 
 // TLOOP:    %[[AB_SUB:.*]] = linalg.tiled_loop (%[[K:.*]]) = (%[[C0]])

mlir/test/Dialect/Linalg/fusion-tensor.mlir

 // RUN: mlir-opt %s -linalg-fusion-for-tensor-ops -split-input-file | FileCheck %s
 
 // CHECK-DAG: [[$MAP0:#[a-zA-Z0-9_]*]] = affine_map<(d0, d1) -> (d0, d1)>
 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 
 // CHECK-LABEL: @add_mul_fusion
 func @add_mul_fusion(%arg0: tensor<?x?xf32>, %arg1 : tensor<?x?xf32>, %arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   %3 = linalg.generic {indexing_maps = [#map0, #map0, #map0], iterator_types = ["parallel", "parallel"]}
       ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%2 : tensor<?x?xf32>) {
     ^bb0(%arg3: f32, %arg4: f32, %arg5: f32):       // no predecessors
       %4 = addf %arg3, %arg4 : f32
       linalg.yield %4 : f32
   } -> tensor<?x?xf32>

 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 #map1 = affine_map<(d0, d1) -> ()>
 
 // CHECK-LABEL: @scalar_add_mul_fusion
 func @scalar_add_mul_fusion(%arg0: tensor<?x?xf32>, %arg1 : f32, %arg2 : f32) -> tensor<?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   %3 = linalg.generic {indexing_maps = [#map0, #map1, #map0], iterator_types = ["parallel", "parallel"]}
       ins(%arg0, %arg1 : tensor<?x?xf32>, f32)
       outs(%2 : tensor<?x?xf32>) {
     ^bb0(%arg3: f32, %arg4: f32, %arg5: f32):       // no predecessors
       %4 = addf %arg3, %arg4 : f32
       linalg.yield %4 : f32
   } -> tensor<?x?xf32>

 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 #map1 = affine_map<(d0, d1) -> (d1, d0)>
 
 // CHECK-LABEL: @transpose_add_mul_fusion
 func @transpose_add_mul_fusion(%arg0: tensor<?x?xf32>, %arg1 : tensor<?x?xf32>, %arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   %3 = linalg.generic {indexing_maps = [#map0, #map1, #map0], iterator_types = ["parallel", "parallel"]}
       ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%2 : tensor<?x?xf32>) {
     ^bb0(%arg3: f32, %arg4: f32, %arg5: f32):       // no predecessors
       %4 = addf %arg3, %arg4 : f32
       linalg.yield %4 : f32
   } -> tensor<?x?xf32>

 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 #map1 = affine_map<(d0, d1) -> (d1, d0)>
 
 // CHECK-LABEL: @add_transpose_mul_fusion
 func @add_transpose_mul_fusion(%arg0: tensor<?x?xf32>, %arg1 : tensor<?x?xf32>, %arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xf32>
   %3 = linalg.generic {indexing_maps = [#map0, #map1, #map0], iterator_types = ["parallel", "parallel"]}
       ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
       outs(%2 : tensor<?x?xf32>) {
     ^bb0(%arg3: f32, %arg4: f32, %arg5: f32):       // no predecessors
       %4 = addf %arg3, %arg4 : f32
       linalg.yield %4 : f32
   } -> tensor<?x?xf32>

 #map1 = affine_map<(d0, d1) -> (d0)>
 #map2 = affine_map<(d0) -> (d0)>
 
 // CHECK-LABEL: @add_broadcast_mul_fusion
 func @add_broadcast_mul_fusion(%arg0: tensor<?xf32>, %arg1 : tensor<?xf32>, %arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?xf32>
   %1 = linalg.init_tensor [%0] : tensor<?xf32>
   %2 = linalg.generic {indexing_maps = [#map2, #map2, #map2], iterator_types = ["parallel"]}
       ins(%arg0, %arg1 : tensor<?xf32>, tensor<?xf32>)
       outs(%1 : tensor<?xf32>) {
     ^bb0(%arg3: f32, %arg4: f32, %arg5: f32):       // no predecessors
       %3 = addf %arg3, %arg4 : f32
       linalg.yield %3 : f32
   } -> tensor<?xf32>
   // CHECK: linalg.generic {
   // CHECK-SAME: indexing_maps = {{\[}}[[$MAP1]], [[$MAP1]], [[$MAP0]], [[$MAP0]]
-  %3 = memref.dim %arg2, %c1 : tensor<?x?xf32>
+  %3 = tensor.dim %arg2, %c1 : tensor<?x?xf32>
   %4 = linalg.init_tensor [%0, %3] : tensor<?x?xf32>
   %5 = linalg.generic {indexing_maps = [#map1, #map0, #map0], iterator_types = ["parallel", "parallel"]}
       ins(%2, %arg2 : tensor<?xf32>, tensor<?x?xf32>)
       outs(%4 : tensor<?x?xf32>){
     ^bb0(%arg5: f32, %arg6: f32, %arg7: f32):       // no predecessors
       %6 = mulf %arg5, %arg6 : f32
       linalg.yield %6 : f32
     } -> tensor<?x?xf32>

 #map0 = affine_map<(d0, d1, d2) -> (d0)>
 #map1 = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
 func @generic_op_constant_fusion(%arg0 : tensor<5x?x?xf32>) -> tensor<5x?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %cst = constant dense<42.0> : tensor<5xf32>
-  %0 = memref.dim %arg0, %c1 : tensor<5x?x?xf32>
-  %1 = memref.dim %arg0, %c2 : tensor<5x?x?xf32>
+  %0 = tensor.dim %arg0, %c1 : tensor<5x?x?xf32>
+  %1 = tensor.dim %arg0, %c2 : tensor<5x?x?xf32>
   %2 = linalg.init_tensor [5, %0, %1] : tensor<5x?x?xf32>
   %3 = linalg.generic {
     indexing_maps = [#map0, #map1, #map1],
     iterator_types = ["parallel", "parallel", "parallel"]}
     ins(%cst, %arg0 : tensor<5xf32>, tensor<5x?x?xf32>)
     outs(%2 : tensor<5x?x?xf32>) {
     ^bb0(%arg1: f32, %arg2: f32, %arg3: f32):
       %4 = mulf %arg1, %arg2 : f32

 #map1 = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
 func @generic_op_zero_dim_constant_fusion(%arg0 : tensor<5x?x?xf32>)
   -> tensor<5x?x?xf32>
 {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %cst = constant dense<42.0> : tensor<f32>
-  %0 = memref.dim %arg0, %c1 : tensor<5x?x?xf32>
-  %1 = memref.dim %arg0, %c2 : tensor<5x?x?xf32>
+  %0 = tensor.dim %arg0, %c1 : tensor<5x?x?xf32>
+  %1 = tensor.dim %arg0, %c2 : tensor<5x?x?xf32>
   %2 = linalg.init_tensor [5, %0, %1] : tensor<5x?x?xf32>
   %3 = linalg.generic {
     indexing_maps = [#map0, #map1, #map1],
     iterator_types = ["parallel", "parallel", "parallel"]}
     ins(%cst, %arg0 : tensor<f32>, tensor<5x?x?xf32>)
     outs(%2 : tensor<5x?x?xf32>) {
     ^bb0(%arg1: f32, %arg2: f32, %arg3: f32):
       %4 = mulf %arg1, %arg2 : f32

 
 // -----
 
 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 func @producer_indexed_consumer_fusion(%arg0: tensor<?x?xi32>,
                                        %arg1: tensor<?x?xi32>) -> tensor<?x?xi32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xi32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xi32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xi32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xi32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xi32>
   %3 = linalg.generic {
     indexing_maps = [#map0, #map0, #map0],
     iterator_types = ["parallel", "parallel"] }
     ins(%arg0, %arg1  : tensor<?x?xi32>, tensor<?x?xi32>)
     outs(%2 : tensor<?x?xi32>) {
     ^bb0(%arg2: i32, %arg3: i32, %arg4: i32):       // no predecessors
       %10 = addi %arg2, %arg3 : i32

 //  CHECK-NOT: linalg.generic
 
 // -----
 
 #map0 = affine_map<(d0, d1) -> (d0, d1)>
 func @indexed_producer_consumer_fusion(%arg0: tensor<?x?xi32>) -> tensor<?x?xi32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xi32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xi32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xi32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xi32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xi32>
   %3 = linalg.generic {
     indexing_maps = [#map0, #map0],
     iterator_types = ["parallel", "parallel"] }
     ins(%arg0 : tensor<?x?xi32>)
     outs(%2 : tensor<?x?xi32>) {
     ^bb0(%arg4: i32, %arg5: i32):       // no predecessors
       %idx0 = linalg.index 0 : index

 
 // The indices of the first generic op are swapped after fusion.
 #map0 = affine_map<(d0, d1) -> (d1, d0)>
 #map1 = affine_map<(d0, d1) -> (d0, d1)>
 func @indexed_producer_indexed_consumer_fusion(%arg0: tensor<?x?xi32>)
                                                -> tensor<?x?xi32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xi32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xi32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xi32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xi32>
   %2 = linalg.init_tensor [%0, %1] : tensor<?x?xi32>
   %3 = linalg.generic {
     indexing_maps = [#map0, #map0],
     iterator_types = ["parallel", "parallel"] }
     ins(%arg0 : tensor<?x?xi32>)
     outs(%2 : tensor<?x?xi32>) {
     ^bb0(%arg2: i32, %arg3: i32):       // no predecessors
       %idx0 = linalg.index 0 : index

 
 #map1 = affine_map<(d0) -> (d0)>
 #map2 = affine_map<(d0, d1) -> (d0, d1)>
 #map3 = affine_map<(d0, d1) -> (d1)>
 func @one_dim_indexed_producer_consumer_fusion(%arg0 : tensor<?xi32>,
                                                %arg1 : tensor<?x?xi32>) -> tensor<?x?xi32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %d0 = memref.dim %arg0, %c0 : tensor<?xi32>
+  %d0 = tensor.dim %arg0, %c0 : tensor<?xi32>
   %0 = linalg.init_tensor [%d0] : tensor<?xi32>
   %1 = linalg.generic
       {indexing_maps = [#map1, #map1],
        iterator_types = ["parallel"]}
       ins(%arg0 : tensor<?xi32>) outs(%0 : tensor<?xi32>) {
       ^bb0(%arg2 : i32, %arg3 : i32):
         %2 = linalg.index 0 : index
         %3 = index_cast %2 : index to i32
         %4 = addi %arg2, %3 : i32
         linalg.yield %4 : i32
       } -> tensor<?xi32>
-  %2 = memref.dim %arg1, %c0 : tensor<?x?xi32>
-  %3 = memref.dim %arg1, %c1 : tensor<?x?xi32>
+  %2 = tensor.dim %arg1, %c0 : tensor<?x?xi32>
+  %3 = tensor.dim %arg1, %c1 : tensor<?x?xi32>
   %4 = linalg.init_tensor [%2, %3] : tensor<?x?xi32>
   %5 = linalg.generic
       {indexing_maps = [#map2, #map3, #map2],
        iterator_types = ["parallel", "parallel"]}
       ins(%arg1, %1 : tensor<?x?xi32>, tensor<?xi32>)
       outs(%4 : tensor<?x?xi32>) {
       ^bb0(%arg2 : i32, %arg3 : i32, %arg4: i32):
         %6 = addi %arg2, %arg3 : i32
func @sigmoid_dynamic_dim(%0: tensor<?x1xf32>) -> tensor<?x1xf32> {
   %1 = linalg.generic {indexing_maps = [
     affine_map<(d0, d1) -> (d0, d1)>],
     iterator_types = ["parallel", "parallel"]
   }
      outs(%init0 : tensor<?x1xf32>) {
     ^bb0(%a: f32):  // no predecessors
       linalg.yield %cp5 : f32
   } -> tensor<?x1xf32>
-  %d0 = memref.dim %0, %c0 : tensor<?x1xf32>
+  %d0 = tensor.dim %0, %c0 : tensor<?x1xf32>
   %init1 = linalg.init_tensor [%d0, 1] : tensor<?x1xf32>
   %2 = linalg.generic {indexing_maps = [
     affine_map<(d0, d1) -> (d0, d1)>,
     affine_map<(d0, d1) -> (d0, d1)>,
     affine_map<(d0, d1) -> (d0, d1)>],
     iterator_types = ["parallel", "parallel"]
   }
       ins(%0, %1 : tensor<?x1xf32>, tensor<?x1xf32>)
%2 = linalg.generic #trait ins(%0 : tensor<?x?xf32>) outs(%0 : tensor<?x?xf32>) {
          linalg.yield %3 : f32
        } -> tensor<?x?xf32>
   return %2 : tensor<?x?xf32>
 }
 //      CHECK: func @break_outs_dependency(
 // CHECK-SAME:   %[[ARG0:.+]]: tensor<?x?xf32>)
 //  CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //  CHECK-DAG:   %[[C1:.+]] = constant 1 : index
-//  CHECK-DAG:   %[[D0:.+]] = memref.dim %[[ARG0]], %[[C0]]
-//  CHECK-DAG:   %[[D1:.+]] = memref.dim %[[ARG0]], %[[C1]]
+//  CHECK-DAG:   %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
+//  CHECK-DAG:   %[[D1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
 //  CHECK-DAG:   %[[INIT:.+]] = linalg.init_tensor [%[[D0]], %[[D1]]]
 //      CHECK:   %[[GENERIC1:.+]] = linalg.generic
 // CHECK-SAME:     outs(%[[INIT]] : tensor<?x?xf32>)
-//  CHECK-DAG:   %[[D0:.+]] = memref.dim %[[GENERIC1]], %[[C0]]
-//  CHECK-DAG:   %[[D1:.+]] = memref.dim %[[GENERIC1]], %[[C1]]
+//  CHECK-DAG:   %[[D0:.+]] = tensor.dim %[[GENERIC1]], %[[C0]]
+//  CHECK-DAG:   %[[D1:.+]] = tensor.dim %[[GENERIC1]], %[[C1]]
 //  CHECK-DAG:   %[[INIT:.+]] = linalg.init_tensor [%[[D0]], %[[D1]]]
 //      CHECK:   %[[RESULT:.+]] = linalg.generic
 // CHECK-SAME:     outs(%[[INIT]] : tensor<?x?xf32>)

mlir/test/Dialect/Linalg/hoist-padding.mlir

func @matmul_tensors(
   %c4 = constant 4 : index
   %cst = constant 0.000000e+00 : f32
 
   //  CHECK-DAG: %[[C0:.*]] = constant 0 : index
   //  CHECK-DAG: %[[C1:.*]] = constant 1 : index
   %c0 = constant 0 : index
   %c1 = constant 1 : index
 
-  //  CHECK-DAG: %[[dM:.*]] = memref.dim %[[TA]], %[[C0]] : tensor<?x?xf32>
-  //  CHECK-DAG: %[[dK:.*]] = memref.dim %[[TA]], %[[C1]] : tensor<?x?xf32>
-  //  CHECK-DAG: %[[dN:.*]] = memref.dim %[[TB]], %[[C1]] : tensor<?x?xf32>
-  %0 = memref.dim %arg0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?xf32>
-  %2 = memref.dim %arg1, %c1 : tensor<?x?xf32>
+  //  CHECK-DAG: %[[dM:.*]] = tensor.dim %[[TA]], %[[C0]] : tensor<?x?xf32>
+  //  CHECK-DAG: %[[dK:.*]] = tensor.dim %[[TA]], %[[C1]] : tensor<?x?xf32>
+  //  CHECK-DAG: %[[dN:.*]] = tensor.dim %[[TB]], %[[C1]] : tensor<?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
+  %2 = tensor.dim %arg1, %c1 : tensor<?x?xf32>
 
   //      CHECK: scf.for %[[I:[0-9a-z]+]] =
   // First padded tensor is MxKx2x4 under loop M so Kx2x4
   //      CHECK:   %[[SZpad0_K:[0-9]+]] = affine.apply #[[$DIVS4]]()[%[[dK]]]
   //      CHECK:   linalg.init_tensor [%[[SZpad0_K]], 2, 4] : tensor<?x2x4xf32>
   // 1-D loop
   //      CHECK:   %[[A:.*]] = scf.for %[[J1:[0-9a-z]+]] =
   // Iteration count along J1
func @matmul_tensors(
   // CHECK-SAME:         tensor<?x?x4x3xf32> to tensor<4x3xf32>
   //      CHECK:       %[[stC:.*]] = linalg.pad_tensor %{{.*}}
   //      CHECK:         : tensor<?x?xf32> to tensor<2x3xf32>
   //      CHECK:       linalg.matmul ins(%[[stA]], %[[stB]] : tensor<2x4xf32>, tensor<4x3xf32>)
   // CHECK-SAME:         outs(%[[stC]] : tensor<2x3xf32>) -> tensor<2x3xf32>
   %3 = scf.for %arg3 = %c0 to %0 step %c2 iter_args(%arg4 = %arg2) -> (tensor<?x?xf32>) {
     %4 = scf.for %arg5 = %c0 to %2 step %c3 iter_args(%arg6 = %arg4) -> (tensor<?x?xf32>) {
       %5 = scf.for %arg7 = %c0 to %1 step %c4 iter_args(%arg8 = %arg6) -> (tensor<?x?xf32>) {
-        %6 = memref.dim %arg0, %c0 : tensor<?x?xf32>
+        %6 = tensor.dim %arg0, %c0 : tensor<?x?xf32>
         %7 = affine.min #map0(%arg3)[%6]
-        %8 = memref.dim %arg0, %c1 : tensor<?x?xf32>
+        %8 = tensor.dim %arg0, %c1 : tensor<?x?xf32>
         %9 = affine.min #map1(%arg7)[%8]
         %10 = tensor.extract_slice %arg0[%arg3, %arg7] [%7, %9] [1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
-        %11 = memref.dim %arg1, %c0 : tensor<?x?xf32>
+        %11 = tensor.dim %arg1, %c0 : tensor<?x?xf32>
         %12 = affine.min #map1(%arg7)[%11]
-        %13 = memref.dim %arg1, %c1 : tensor<?x?xf32>
+        %13 = tensor.dim %arg1, %c1 : tensor<?x?xf32>
         %14 = affine.min #map2(%arg5)[%13]
         %15 = tensor.extract_slice %arg1[%arg7, %arg5] [%12, %14] [1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
-        %16 = memref.dim %arg8, %c0 : tensor<?x?xf32>
+        %16 = tensor.dim %arg8, %c0 : tensor<?x?xf32>
         %17 = affine.min #map3(%16, %arg3)
-        %18 = memref.dim %arg8, %c1 : tensor<?x?xf32>
+        %18 = tensor.dim %arg8, %c1 : tensor<?x?xf32>
         %19 = affine.min #map4(%18, %arg5)
         %20 = tensor.extract_slice %arg8[%arg3, %arg5] [%17, %19] [1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
         %21 = subi %c2, %7 : index
         %22 = subi %c4, %9 : index
         %23 = linalg.pad_tensor %10 low[%c0, %c0] high[%21, %22] {
         ^bb0(%arg9: index, %arg10: index):  // no predecessors
           linalg.yield %cst : f32
         } : tensor<?x?xf32> to tensor<2x4xf32>

 func @dot(%arg0: tensor<?xf32>, %arg1: tensor<?xf32>, %arg2: tensor<f32>)
     -> tensor<f32>
 {
   %c8 = constant 8 : index
   %c4 = constant 4 : index
   %cst = constant 0.000000e+00 : f32
   %c2 = constant 2 : index
   %c0 = constant 0 : index
-  %1 = memref.dim %arg0, %c0 : tensor<?xf32>
-  %2 = memref.dim %arg0, %c0 : tensor<?xf32>
-  %3 = memref.dim %arg1, %c0 : tensor<?xf32>
+  %1 = tensor.dim %arg0, %c0 : tensor<?xf32>
+  %2 = tensor.dim %arg0, %c0 : tensor<?xf32>
+  %3 = tensor.dim %arg1, %c0 : tensor<?xf32>
 
   //      CHECK: scf.for %[[I:[0-9a-z]+]] =
   //
   //      CHECK:   %[[MR8:.*]] = affine.min #[[$MIN_REST8]](%[[I]])
   //      CHECK:   %[[D0:.*]] = affine.apply #[[$DIV4]](%[[MR8]])
   //      CHECK:   %[[MM4:.*]] = affine.min #[[$MIN_MOD4]](%[[MR8]])
   //      CHECK:   %[[D1:.*]] = affine.apply #[[$DIV2]](%[[MM4]])
   // Init tensor and pack.

mlir/test/Dialect/Linalg/reshape_fusion.mlir

 // -----
 
 func @unit_dim_reshape_expansion_full
   (%arg0 : tensor<1x?x1x2x1x4xf32>, %arg1 : tensor<?x2x4xf32>)
   -> tensor<?x2x4xf32> {
   %c1 = constant 1 : index
   %0 = linalg.tensor_collapse_shape %arg0 [[0, 1, 2], [3, 4], [5]]
     : tensor<1x?x1x2x1x4xf32> into tensor<?x2x4xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<1x?x1x2x1x4xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<1x?x1x2x1x4xf32>
   %2 = linalg.init_tensor [%1, 2, 4] : tensor<?x2x4xf32>
   %3 = linalg.generic
     {indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d1, d2)>,
                       affine_map<(d0, d1, d2) -> (d0, d1, d2)>,
                       affine_map<(d0, d1, d2) -> (d0, d1, d2)>],
      iterator_types = ["parallel", "parallel", "parallel"]}
     ins(%0, %arg1 : tensor<?x2x4xf32>, tensor<?x2x4xf32>)
     outs(%2 : tensor<?x2x4xf32>) {

mlir/test/Dialect/Linalg/resolve-shaped-type-result-dims.mlir

 // RUN: mlir-opt -resolve-shaped-type-result-dims -split-input-file %s | FileCheck %s
 
 func @init_tensor_static_dim() -> (index, index) {
   %c0 = constant 0 : index
   %c2 = constant 2 : index
   %c6 = constant 6 : index
   %0 = linalg.init_tensor [4, 5, %c6] : tensor<4x5x?xf32>
-  %1 = memref.dim %0, %c2 : tensor<4x5x?xf32>
-  %2 = memref.dim %0, %c0 : tensor<4x5x?xf32>
+  %1 = tensor.dim %0, %c2 : tensor<4x5x?xf32>
+  %2 = tensor.dim %0, %c0 : tensor<4x5x?xf32>
   return %1, %2 : index, index
 }
 //      CHECK: func @init_tensor_static_dim
 //  CHECK-DAG:   %[[C4:.+]] = constant 4 : index
 //  CHECK-DAG:   %[[C6:.+]] = constant 6 : index
 //      CHECK:   return %[[C6]], %[[C4]]
 
 // -----
 
 func @init_tensor_dynamic_dim(%arg0 : index) -> (index) {
   %c2 = constant 2 : index
   %0 = linalg.init_tensor [4, 5, %arg0] : tensor<4x5x?xf32>
-  %1 = memref.dim %0, %c2 : tensor<4x5x?xf32>
+  %1 = tensor.dim %0, %c2 : tensor<4x5x?xf32>
   return %1 : index
 }
 //      CHECK: func @init_tensor_dynamic_dim
 // CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: index
 //      CHECK:   return %[[ARG0]]
 
 // -----
 
 func @init_tensor_dynamic_dim2(%arg0 : index, %arg1 : index) -> (index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %0 = linalg.init_tensor [%arg0, %arg1] : tensor<?x?xf32>
-  %1 = memref.dim %0, %c0 : tensor<?x?xf32>
-  %2 = memref.dim %0, %c1 : tensor<?x?xf32>
+  %1 = tensor.dim %0, %c0 : tensor<?x?xf32>
+  %2 = tensor.dim %0, %c1 : tensor<?x?xf32>
   return %1, %2 : index, index
 }
 //      CHECK: func @init_tensor_dynamic_dim2
 // CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: index
 // CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: index
 //      CHECK:   return %[[ARG0]], %[[ARG1]]
 
 // -----
%0 = linalg.generic
      iterator_types = ["parallel", "parallel", "reduction"]}
     ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
     outs(%arg2 : tensor<?x?xf32>) {
     ^bb0(%arg3 : f32, %arg4 : f32, %arg5 : f32):
       %1 = mulf %arg3, %arg4 : f32
       %2 = addf %1, %arg5 : f32
       linalg.yield %2 : f32
     } -> tensor<?x?xf32>
-  %3 = memref.dim %0, %c0 : tensor<?x?xf32>
-  %4 = memref.dim %0, %c1 : tensor<?x?xf32>
+  %3 = tensor.dim %0, %c0 : tensor<?x?xf32>
+  %4 = tensor.dim %0, %c1 : tensor<?x?xf32>
   return %3, %4 : index, index
 }
 //       CHECK: #[[MAP0:.+]] = affine_map<()[s0, s1] -> (s0 + s1)>
 //       CHECK: #[[MAP1:.+]] = affine_map<()[s0, s1] -> (s1 - s0)>
 //       CHECK: func @remove_dim_result_uses
 //  CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C1:.+]] = constant 1 : index
-//   CHECK-DAG:   %[[T0:.+]] = memref.dim %[[ARG0]], %[[C0]]
-//   CHECK-DAG:   %[[T1:.+]] = memref.dim %[[ARG1]], %[[C1]]
+//   CHECK-DAG:   %[[T0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
+//   CHECK-DAG:   %[[T1:.+]] = tensor.dim %[[ARG1]], %[[C1]]
 //       CHECK:   %[[T2:.+]] = affine.apply #[[MAP0]]()[%[[T0]], %[[T1]]]
-//   CHECK-DAG:   %[[T3:.+]] = memref.dim %[[ARG0]], %[[C0]]
-//   CHECK-DAG:   %[[T4:.+]] = memref.dim %[[ARG1]], %[[C1]]
+//   CHECK-DAG:   %[[T3:.+]] = tensor.dim %[[ARG0]], %[[C0]]
+//   CHECK-DAG:   %[[T4:.+]] = tensor.dim %[[ARG1]], %[[C1]]
 //       CHECK:   %[[T5:.+]] = affine.apply #[[MAP1]]()[%[[T3]], %[[T4]]]
 //       CHECK:   return %[[T2]], %[[T5]]
 
 // -----
 
 func @remove_dim_result_uses_outs
   (%arg0 : tensor<?xf32>, %arg1 : index) -> (index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %d0 = memref.dim %arg0, %c0 : tensor<?xf32>
+  %d0 = tensor.dim %arg0, %c0 : tensor<?xf32>
   %0 = linalg.init_tensor [%d0, %arg1] : tensor<?x?xf32>
   %1 = linalg.generic
     {indexing_maps = [affine_map<(d0, d1) -> (d0)>,
                       affine_map<(d0, d1) -> (d0, d1)>],
      iterator_types = ["parallel", "parallel"]}
     ins(%arg0 : tensor<?xf32>) outs(%0 : tensor<?x?xf32>) {
     ^bb0(%arg2: f32, %arg3: f32) :
       linalg.yield %arg2 : f32
     } -> tensor<?x?xf32>
-  %2 = memref.dim %1, %c1 : tensor<?x?xf32>
+  %2 = tensor.dim %1, %c1 : tensor<?x?xf32>
   return %2 : index
 }
 //      CHECK: func @remove_dim_result_uses_outs
 // CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: index
 //      CHECK:   return %[[ARG1]]
 
 // -----
 
 func @remove_dim_result_uses_sequence
   (%arg0 : tensor<?x?xf32>, %arg1 : tensor<?x?xf32>,
    %arg2 : tensor<?x?xf32>) -> (index, index, index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %0 = linalg.matmul ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
     outs(%arg2 : tensor<?x?xf32>) -> tensor<?x?xf32>
-  %1 = memref.dim %0, %c0 : tensor<?x?xf32>
-  %2 = memref.dim %0, %c1 : tensor<?x?xf32>
+  %1 = tensor.dim %0, %c0 : tensor<?x?xf32>
+  %2 = tensor.dim %0, %c1 : tensor<?x?xf32>
   %3 = linalg.generic
     {indexing_maps = [affine_map<(d0, d1, d2) -> (d1, d0)>,
                       affine_map<(d0, d1, d2) -> (d0, d2)>,
                       affine_map<(d0, d1, d2) -> (d0, d2)>],
      iterator_types = ["parallel", "reduction", "parallel"]}
     ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>)
     outs(%0 : tensor<?x?xf32>) {
     ^bb0(%arg3 : f32, %arg4 : f32, %arg5 : f32):
       %4 = mulf %arg3, %arg4 : f32
       %5 = addf %4, %arg5 : f32
       linalg.yield %5 : f32
     } -> tensor<?x?xf32>
-  %6 = memref.dim %3, %c0 : tensor<?x?xf32>
-  %7 = memref.dim %3, %c1 : tensor<?x?xf32>
+  %6 = tensor.dim %3, %c0 : tensor<?x?xf32>
+  %7 = tensor.dim %3, %c1 : tensor<?x?xf32>
   return %1, %2, %6, %7 : index, index, index, index
 }
 // CHECK-LABEL: func @remove_dim_result_uses_sequence
 //  CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //  CHECK-SAME:   %[[ARG2:[a-zA-Z0-9_]+]]: tensor<?x?xf32>
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C1:.+]] = constant 1 : index
-//   CHECK-DAG:   %[[T0:.+]] = memref.dim %[[ARG0]], %[[C0]]
-//   CHECK-DAG:   %[[T1:.+]] = memref.dim %[[ARG1]], %[[C1]]
-//   CHECK-DAG:   %[[T2:.+]] = memref.dim %[[ARG0]], %[[C1]]
-//   CHECK-DAG:   %[[T3:.+]] = memref.dim %[[ARG1]], %[[C1]]
+//   CHECK-DAG:   %[[T0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
+//   CHECK-DAG:   %[[T1:.+]] = tensor.dim %[[ARG1]], %[[C1]]
+//   CHECK-DAG:   %[[T2:.+]] = tensor.dim %[[ARG0]], %[[C1]]
+//   CHECK-DAG:   %[[T3:.+]] = tensor.dim %[[ARG1]], %[[C1]]
 //       CHECK:   return %[[T0]], %[[T1]], %[[T2]], %[[T3]]
 
 // -----
 
 func @keep_result_dim_uses_sequence2
   (%arg0 : tensor<?xf32>, %arg1 : index) -> (index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
-  %d0 = memref.dim %arg0, %c0 : tensor<?xf32>
+  %d0 = tensor.dim %arg0, %c0 : tensor<?xf32>
   %0 = linalg.init_tensor [%d0, %arg1] : tensor<?x?xf32>
   %1 = linalg.generic
     {indexing_maps = [affine_map<(d0, d1) -> (d0)>,
                       affine_map<(d0, d1) -> (d0, d1)>],
      iterator_types = ["parallel", "parallel"]}
     ins(%arg0 : tensor<?xf32>) outs(%0 : tensor<?x?xf32>) {
     ^bb0(%arg2: f32, %arg3 : f32):
       linalg.yield %arg2 : f32
     } -> tensor<?x?xf32>
-  %2 = memref.dim %1, %c0 : tensor<?x?xf32>
-  %3 = memref.dim %1, %c1 : tensor<?x?xf32>
+  %2 = tensor.dim %1, %c0 : tensor<?x?xf32>
+  %3 = tensor.dim %1, %c1 : tensor<?x?xf32>
   return %2, %3 : index, index
 }
 //       CHECK: func @keep_result_dim_uses_sequence2
 //  CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<?xf32>
 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: index
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
-//   CHECK-DAG:   %[[T0:.+]] = memref.dim %[[ARG0]], %[[C0]]
+//   CHECK-DAG:   %[[T0:.+]] = tensor.dim %[[ARG0]], %[[C0]]
 //       CHECK:   return %[[T0]], %[[ARG1]]
 
 // -----
 
 #map = affine_map<(d0) -> (d0)>
 
 func @init_tensor_dim_of_linalg_result(%arg_0 : tensor<?xf32>,
     %arg_1: tensor<?xf32>) -> (index, index) {
   %0, %1 = linalg.generic {
     indexing_maps = [#map, #map, #map],
     iterator_types = ["parallel"]
   } ins(%arg_0 : tensor<?xf32>)
     outs(%arg_0, %arg_1 : tensor<?xf32>, tensor<?xf32>) {
   ^bb0(%in: f32, %out_0: f32, %out_1: f32):
     linalg.yield %in, %in : f32, f32
   } -> (tensor<?xf32>, tensor<?xf32>)
 
   %c0 = constant 0 : index
-  %num_elem_0 = memref.dim %0, %c0 : tensor<?xf32>
+  %num_elem_0 = tensor.dim %0, %c0 : tensor<?xf32>
 
-  %num_elem_1 = memref.dim %1, %c0 : tensor<?xf32>
+  %num_elem_1 = tensor.dim %1, %c0 : tensor<?xf32>
   return %num_elem_0, %num_elem_1 : index, index
 }
 //      CHECK: func @init_tensor_dim_of_linalg_result(
 // CHECK-SAME:   %[[ARG_0:[a-zA-Z0-9_]+]]: tensor<?xf32>
 // CHECK-SAME:   %[[ARG_1:[a-zA-Z0-9_]+]]: tensor<?xf32>)
-//      CHECK:   %[[R0:.+]] = memref.dim %[[ARG_0]]
-//      CHECK:   %[[R1:.+]] = memref.dim %[[ARG_0]]
+//      CHECK:   %[[R0:.+]] = tensor.dim %[[ARG_0]]
+//      CHECK:   %[[R1:.+]] = tensor.dim %[[ARG_0]]
 //      CHECK:   return %[[R0]], %[[R1]]
 
 // -----
 
 func @dim_reshape_expansion(%arg0 : tensor<6x5x?xf32>) -> (index, index, index)
 {
   %c1 = constant 1 : index
   %c3 = constant 3 : index
   %c4 = constant 4 : index
   %0 = linalg.tensor_expand_shape %arg0 [[0, 1], [2], [3, 4, 5]]
       : tensor<6x5x?xf32> into tensor<2x3x5x4x?x7xf32>
-  %1 = memref.dim %0, %c1 : tensor<2x3x5x4x?x7xf32>
-  %2 = memref.dim %0, %c3 : tensor<2x3x5x4x?x7xf32>
-  %3 = memref.dim %0, %c4 : tensor<2x3x5x4x?x7xf32>
+  %1 = tensor.dim %0, %c1 : tensor<2x3x5x4x?x7xf32>
+  %2 = tensor.dim %0, %c3 : tensor<2x3x5x4x?x7xf32>
+  %3 = tensor.dim %0, %c4 : tensor<2x3x5x4x?x7xf32>
   return %1, %2, %3 : index, index, index
 }
 //      CHECK: #[[MAP:.+]] = affine_map<()[s0] -> (s0 floordiv 28)>
 //      CHECK: func @dim_reshape_expansion
 // CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<6x5x?xf32>
 //  CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //  CHECK-DAG:   %[[C3:.+]] = constant 3 : index
 //  CHECK-DAG:   %[[C4:.+]] = constant 4 : index
-//      CHECK:   %[[D0:.+]] = memref.dim %[[ARG0]], %[[C2]]
+//      CHECK:   %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C2]]
 //      CHECK:   %[[D1:.+]] = affine.apply #[[MAP]]()[%[[D0]]]
 //      CHECK:   return %[[C3]], %[[C4]], %[[D1]]
 
 // -----
 
 func @dim_reshape_collapse(%arg0 : tensor<2x3x5x4x?x7xf32>) -> (index, index)
 {
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %0 = linalg.tensor_collapse_shape %arg0 [[0, 1], [2], [3, 4, 5]]
       : tensor<2x3x5x4x?x7xf32> into tensor<6x5x?xf32>
-  %1 = memref.dim %0, %c1 : tensor<6x5x?xf32>
-  %2 = memref.dim %0, %c2 : tensor<6x5x?xf32>
+  %1 = tensor.dim %0, %c1 : tensor<6x5x?xf32>
+  %2 = tensor.dim %0, %c2 : tensor<6x5x?xf32>
   return %1, %2 : index, index
 }
 //      CHECK: #[[MAP:.+]] = affine_map<()[s0] -> (s0 * 28)>
 //      CHECK: func @dim_reshape_collapse
 // CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: tensor<2x3x5x4x?x7xf32>
 //  CHECK-DAG:   %[[C4:.+]] = constant 4 : index
 //  CHECK-DAG:   %[[C5:.+]] = constant 5 : index
-//      CHECK:   %[[D0:.+]] = memref.dim %[[ARG0]], %[[C4]]
+//      CHECK:   %[[D0:.+]] = tensor.dim %[[ARG0]], %[[C4]]
 //      CHECK:   %[[D1:.+]] = affine.apply #[[MAP]]()[%[[D0]]]
 //      CHECK:   return %[[C5]], %[[D1]]
 
 // -----
 
 func @dim_of_pad_op(%arg0 : tensor<2x?x?xf32>, %arg1 : index, %arg2 : index,
     %arg3: f32) -> (index, index, index)
 {
    %c0 = constant 0 : index
    %c1 = constant 1 : index
    %c2 = constant 2 : index
    %c3 = constant 3 : index
    %c4 = constant 4 : index
    %c5 = constant 5 : index
    %0 = linalg.pad_tensor %arg0 low[%c3, %arg1, %c4] high[7, %c5, %arg2] {
      ^bb0(%arg4: index, %arg5: index, %arg6: index):
        linalg.yield %arg3 : f32
    } : tensor<2x?x?xf32> to tensor<?x?x?xf32>
-   %1 = memref.dim %0, %c0 : tensor<?x?x?xf32>
-   %2 = memref.dim %0, %c1 : tensor<?x?x?xf32>
-   %3 = memref.dim %0, %c2 : tensor<?x?x?xf32>
+   %1 = tensor.dim %0, %c0 : tensor<?x?x?xf32>
+   %2 = tensor.dim %0, %c1 : tensor<?x?x?xf32>
+   %3 = tensor.dim %0, %c2 : tensor<?x?x?xf32>
    return %1, %2, %3 : index, index, index
 }
 //  CHECK-DAG: #[[MAP0:.+]] = affine_map<()[s0, s1] -> (s1 + s0 + 5)>
 //  CHECK-DAG: #[[MAP1:.+]] = affine_map<()[s0, s1] -> (s1 + s0 + 4)>
 //      CHECK: func @dim_of_pad_op
 // CHECK-SAME:   %[[ARG0:[A-Za-z0-9_]+]]: tensor<2x?x?xf32>
 // CHECK-SAME:   %[[ARG1:[A-Za-z0-9_]+]]: index
 // CHECK-SAME:   %[[ARG2:[A-Za-z0-9_]+]]: index
 //  CHECK-DAG:   %[[C1:.+]] = constant 1 : index
 //  CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //  CHECK-DAG:   %[[C12:.+]] = constant 12 : index
-//      CHECK:   %[[IN_DIM1:.+]] = memref.dim %[[ARG0]], %[[C1]]
+//      CHECK:   %[[IN_DIM1:.+]] = tensor.dim %[[ARG0]], %[[C1]]
 //      CHECK:   %[[OUT_DIM1:.+]] = affine.apply #[[MAP0]]()[%[[ARG1]], %[[IN_DIM1]]]
-//      CHECK:   %[[IN_DIM2:.+]] = memref.dim %[[ARG0]], %[[C2]]
+//      CHECK:   %[[IN_DIM2:.+]] = tensor.dim %[[ARG0]], %[[C2]]
 //      CHECK:   %[[OUT_DIM2:.+]] = affine.apply #[[MAP1]]()[%[[ARG2]], %[[IN_DIM2]]]
 //      CHECK:   return %[[C12]], %[[OUT_DIM1]], %[[OUT_DIM2]]

mlir/test/Dialect/Linalg/roundtrip.mlir

func @tiled_loop_reduction(%input_3d: tensor<16x24x32xf32>,
                            %input_2d: tensor<16x32xf32>,
                            %input_1d: tensor<24xf32>,
                            %output: tensor<24xf32>) -> tensor<24xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %c4 = constant 4 : index
   %c8 = constant 8 : index
-  %X = memref.dim %input_3d, %c0 : tensor<16x24x32xf32>
-  %Y = memref.dim %input_3d, %c1 : tensor<16x24x32xf32>
-  %Z = memref.dim %input_3d, %c2 : tensor<16x24x32xf32>
+  %X = tensor.dim %input_3d, %c0 : tensor<16x24x32xf32>
+  %Y = tensor.dim %input_3d, %c1 : tensor<16x24x32xf32>
+  %Z = tensor.dim %input_3d, %c2 : tensor<16x24x32xf32>
   %result = linalg.tiled_loop (%i, %j, %k)
       = (%c0, %c0, %c0) to (%X, %Y, %Z) step (%c2, %c4, %c8)
       ins(%i3d_ = %input_3d: tensor<16x24x32xf32>,
           %i2d_ = %input_2d: tensor<16x32xf32>,
           %i1d_ = %input_1d: tensor<24xf32>)
       outs(%o_ =  %output: tensor<24xf32>)
       iterators["reduction", "parallel", "reduction"]
       distribution["block_x", "block_y", "none"] {

mlir/test/Dialect/Linalg/subtensor-of-padtensor.mlir

 }
 
 // -----
 
 // CHECK-LABEL: @dynamic_high_pad
 //  CHECK-SAME:     %[[ARG0:.*]]: tensor<?x5xf32>
 //   CHECK-NOT:   linalg.pad_tensor
 //       CHECK:   %[[C0:.*]] = constant 0 : index
-//       CHECK:   memref.dim %[[ARG0]], %[[C0]]
+//       CHECK:   tensor.dim %[[ARG0]], %[[C0]]
 //       CHECK:   %[[RESULT:.*]] = scf.if %{{.*}} -> (tensor<3x4xf32>) {
 //       CHECK:     %[[GEN:.*]] = tensor.generate
 //       CHECK:     scf.yield %[[GEN]]
 //       CHECK:   } else {
 //       CHECK:     %[[SUBTENSOR:.*]] = tensor.extract_slice %[[ARG0]][%{{.*}}, 4] [%{{.*}}, 1] [1, 1] : tensor<?x5xf32> to tensor<?x1xf32>
 //       CHECK:     %[[PADTENSOR:.*]] = linalg.pad_tensor %[[SUBTENSOR]] low[0, 0] high[%{{.*}}, 3]
 //       CHECK:     %[[CAST:.*]] = tensor.cast %[[PADTENSOR]] : tensor<?x4xf32> to tensor<3x4xf32>
 //       CHECK:     scf.yield %[[CAST]]

mlir/test/Dialect/Linalg/tile-and-fuse-tensors.mlir

 // RUN: mlir-opt %s -test-linalg-greedy-fusion -split-input-file | FileCheck %s
 
 func @matmul_tensors(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32>, %arg2: tensor<?x?xf32>) -> tensor<?x?xf32> {
   %t0 = linalg.matmul ins(%arg0, %arg1: tensor<?x?xf32>, tensor<?x?xf32>)
                      outs(%arg2: tensor<?x?xf32>)
     -> tensor<?x?xf32>
 
   %c4 = constant 4 : index
   %c2 = constant 2 : index
   %c0 = constant 0 : index
   %c3 = constant 3 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %t0, %c0 : tensor<?x?xf32>
-  %1 = memref.dim %t0, %c1 : tensor<?x?xf32>
-  %2 = memref.dim %arg1, %c1 : tensor<?x?xf32>
+  %0 = tensor.dim %t0, %c0 : tensor<?x?xf32>
+  %1 = tensor.dim %t0, %c1 : tensor<?x?xf32>
+  %2 = tensor.dim %arg1, %c1 : tensor<?x?xf32>
   %3 = scf.for %arg3 = %c0 to %0 step %c2 iter_args(%arg4 = %arg2) -> (tensor<?x?xf32>) {
     %4 = scf.for %arg5 = %c0 to %2 step %c3 iter_args(%arg6 = %arg4) -> (tensor<?x?xf32>) {
       %5 = scf.for %arg7 = %c0 to %1 step %c4 iter_args(%arg8 = %arg6) -> (tensor<?x?xf32>) {
         %6 = tensor.extract_slice %t0[%arg3, %arg7][%c2, 4][1, 1] : tensor<?x?xf32> to tensor<?x4xf32>
         %7 = tensor.extract_slice %arg1[%arg7, %arg5][4, %c3][1, 1] : tensor<?x?xf32> to tensor<4x?xf32>
         %8 = tensor.extract_slice %arg8[%arg3, %arg5][%c2, %c3][1, 1] : tensor<?x?xf32> to tensor<?x?xf32>
         %9 = linalg.matmul ins(%6, %7 : tensor<?x4xf32>, tensor<4x?xf32>) outs(%8 : tensor<?x?xf32>) -> tensor<?x?xf32>
         %10 = tensor.insert_slice %9 into %arg8[%arg3, %arg5] [%c2, %c3] [1, 1]  : tensor<?x?xf32> into tensor<?x?xf32>

 
 //       CHECK: func @matmul_tensors(
 //  CHECK-SAME: %[[A:[0-9a-z]*]]: tensor<?x?xf32>
 //  CHECK-SAME: %[[B:[0-9a-z]*]]: tensor<?x?xf32>
 //  CHECK-SAME: %[[C:[0-9a-z]*]]: tensor<?x?xf32>
 
 //   CHECK-DAG: %[[C0:.*]] = constant 0 : index
 //   CHECK-DAG: %[[C1:.*]] = constant 1 : index
-//   CHECK-DAG: %[[dA0:.*]] = memref.dim %[[A]], %[[C0]] : tensor<?x?xf32>
-//   CHECK-DAG: %[[dA1:.*]] = memref.dim %[[A]], %[[C1]] : tensor<?x?xf32>
-//   CHECK-DAG: %[[dB0:.*]] = memref.dim %[[B]], %[[C0]] : tensor<?x?xf32>
-//   CHECK-DAG: %[[dB1:.*]] = memref.dim %[[B]], %[[C1]] : tensor<?x?xf32>
-//   CHECK-DAG: %[[dC0:.*]] = memref.dim %[[C]], %[[C0]] : tensor<?x?xf32>
-//   CHECK-DAG: %[[dC1:.*]] = memref.dim %[[C]], %[[C1]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dA0:.*]] = tensor.dim %[[A]], %[[C0]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dA1:.*]] = tensor.dim %[[A]], %[[C1]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dB0:.*]] = tensor.dim %[[B]], %[[C0]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dB1:.*]] = tensor.dim %[[B]], %[[C1]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dC0:.*]] = tensor.dim %[[C]], %[[C0]] : tensor<?x?xf32>
+//   CHECK-DAG: %[[dC1:.*]] = tensor.dim %[[C]], %[[C1]] : tensor<?x?xf32>
 //       CHECK: scf.for %[[I:[0-9a-z]*]]
 //       CHECK:   %[[sizeA0:.*]] = affine.min #[[BOUND2_MAP]](%[[I]])[%[[dA0]]]
 //       CHECK:   %[[stA:.*]] = tensor.extract_slice %[[A]][%[[I]], 0] [%[[sizeA0]], %[[dA1]]] [1, 1]  : tensor<?x?xf32> to tensor<?x?xf32>
 //       CHECK:   %[[sizeC0:.*]] = affine.min #[[BOUND2_MAP]](%[[I]])[%[[dC0]]]
 //  CHECK-NEXT:   scf.for %[[J:[0-9a-z]*]]
 //  CHECK-NEXT:     scf.for %[[K:[0-9a-z]*]] {{.*}} iter_args(%[[RES:[0-9a-z]*]]
 //   CHECK-DAG:       %[[stB1:.*]] = tensor.extract_slice %[[B]][%[[K]], %[[J]]] [4, 3] [1, 1]  : tensor<?x?xf32> to tensor<4x3xf32>
 //   CHECK-DAG:       %[[stF:.*]] = tensor.extract_slice %[[RES]][%[[I]], %[[J]]] [2, 3] [1, 1]  : tensor<?x?xf32> to tensor<2x3xf32>
func @conv_tensors_dynamic(%input: tensor<?x?x?x?xf32>, %filter: tensor<?x?x?x?xf32>, %elementwise: tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %c3 = constant 3 : index
   %c4 = constant 4 : index
   %c8 = constant 8 : index
   %c16 = constant 16 : index
 
-  %n = memref.dim %elementwise, %c0 : tensor<?x?x?x?xf32>
-  %oh = memref.dim %elementwise, %c1 : tensor<?x?x?x?xf32>
-  %ow = memref.dim %elementwise, %c2 : tensor<?x?x?x?xf32>
-  %oc = memref.dim %elementwise, %c3 : tensor<?x?x?x?xf32>
+  %n = tensor.dim %elementwise, %c0 : tensor<?x?x?x?xf32>
+  %oh = tensor.dim %elementwise, %c1 : tensor<?x?x?x?xf32>
+  %ow = tensor.dim %elementwise, %c2 : tensor<?x?x?x?xf32>
+  %oc = tensor.dim %elementwise, %c3 : tensor<?x?x?x?xf32>
 
   %init = linalg.init_tensor [%n, %oh, %ow, %oc] : tensor<?x?x?x?xf32>
   %fill = linalg.fill(%cst, %init) : f32, tensor<?x?x?x?xf32> -> tensor<?x?x?x?xf32>
 
   %conv = linalg.conv_2d_input_nhwc_filter_hwcf
     {dilations = dense<1> : tensor<2xi64>, strides = dense<2> : tensor<2xi64>}
     ins(%input, %filter : tensor<?x?x?x?xf32>, tensor<?x?x?x?xf32>)
     outs(%fill : tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>

 //      CHECK: func @conv_tensors_dynamic
 // CHECK-SAME: (%[[INPUT]]: tensor<?x?x?x?xf32>, %[[FILTER]]: tensor<?x?x?x?xf32>, %[[ELEM]]: tensor<?x?x?x?xf32>)
 
 //  CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //  CHECK-DAG:   %[[C1:.+]] = constant 1 : index
 //  CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //  CHECK-DAG:   %[[C3:.+]] = constant 3 : index
 
-//  CHECK-DAG:   %[[ELEM_N:.+]] = memref.dim %[[ELEM]], %[[C0]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[ELEM_OH:.+]] = memref.dim %[[ELEM]], %[[C1]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[ELEM_OW:.+]] = memref.dim %[[ELEM]], %[[C2]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[ELEM_OC:.+]] = memref.dim %[[ELEM]], %[[C3]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[ELEM_N:.+]] = tensor.dim %[[ELEM]], %[[C0]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[ELEM_OH:.+]] = tensor.dim %[[ELEM]], %[[C1]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[ELEM_OW:.+]] = tensor.dim %[[ELEM]], %[[C2]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[ELEM_OC:.+]] = tensor.dim %[[ELEM]], %[[C3]] : tensor<?x?x?x?xf32>
 
 //      CHECK:   %[[INIT:.+]] = linalg.init_tensor [%[[ELEM_N]], %[[ELEM_OH]], %[[ELEM_OW]], %[[ELEM_OC]]] : tensor<?x?x?x?xf32>
 //      CHECK:   %[[FILL:.+]] = linalg.fill(%cst, %[[INIT]]) : f32, tensor<?x?x?x?xf32> -> tensor<?x?x?x?xf32>
 
-//  CHECK-DAG:   %[[FILTER_H:.+]] = memref.dim %[[FILTER]], %[[C0]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILTER_W:.+]] = memref.dim %[[FILTER]], %[[C1]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[INPUT_N:.+]] = memref.dim %[[INPUT]], %[[C0]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[INPUT_H:.+]] = memref.dim %[[INPUT]], %[[C1]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[INPUT_W:.+]] = memref.dim %[[INPUT]], %[[C2]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[INPUT_C:.+]] = memref.dim %[[INPUT]], %[[C3]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILTER_IC:.+]] = memref.dim %[[FILTER]], %[[C2]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILTER_OC:.+]] = memref.dim %[[FILTER]], %[[C3]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILL_N:.+]] = memref.dim %[[FILL]], %[[C0]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILL_H:.+]] = memref.dim %[[FILL]], %[[C1]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILL_W:.+]] = memref.dim %[[FILL]], %[[C2]] : tensor<?x?x?x?xf32>
-//  CHECK-DAG:   %[[FILL_C:.+]] = memref.dim %[[FILL]], %[[C3]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILTER_H:.+]] = tensor.dim %[[FILTER]], %[[C0]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILTER_W:.+]] = tensor.dim %[[FILTER]], %[[C1]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[INPUT_N:.+]] = tensor.dim %[[INPUT]], %[[C0]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[INPUT_H:.+]] = tensor.dim %[[INPUT]], %[[C1]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[INPUT_W:.+]] = tensor.dim %[[INPUT]], %[[C2]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[INPUT_C:.+]] = tensor.dim %[[INPUT]], %[[C3]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILTER_IC:.+]] = tensor.dim %[[FILTER]], %[[C2]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILTER_OC:.+]] = tensor.dim %[[FILTER]], %[[C3]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILL_N:.+]] = tensor.dim %[[FILL]], %[[C0]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILL_H:.+]] = tensor.dim %[[FILL]], %[[C1]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILL_W:.+]] = tensor.dim %[[FILL]], %[[C2]] : tensor<?x?x?x?xf32>
+//  CHECK-DAG:   %[[FILL_C:.+]] = tensor.dim %[[FILL]], %[[C3]] : tensor<?x?x?x?xf32>
 
 //      CHECK:   scf.for %[[IV0:.+]] = %{{.+}} to %[[ELEM_N]] step %{{.+}} iter_args(%{{.+}} = %[[FILL]])
 // CHECK-NEXT:     %[[SIZE_ELEM_N:.+]] = affine.min #[[BOUND8_MAP]](%[[IV0]])[%[[ELEM_N]]]
 // CHECK-NEXT:     %[[SIZE_INPUT_N:.+]] = affine.min #[[BOUND8_MAP_2]](%[[IV0]])[%[[INPUT_N]], %[[ELEM_N]]]
 // CHECK-NEXT:     %[[SIZE_ELEM_N_2:.+]] = affine.min #[[BOUND8_MAP_2]](%[[IV0]])[%[[FILL_N]], %[[ELEM_N]]]
 // CHECK-NEXT:     scf.for %[[IV1:.+]] = %{{.+}} to %[[ELEM_OH]]
 // CHECK-NEXT:       %[[SIZE_ELEM_OH:.+]] = affine.min #[[BOUND16_MAP]](%[[IV1]])[%[[ELEM_OH]]]
 // CHECK-NEXT:       %[[OFFSET_OH:.+]] = affine.apply #[[X2_MAP]](%[[IV1]])

 //     CHECK:       tensor.insert_slice
 func @pad_generic_static(%small_input: tensor<58x1xf32>, %large_input: tensor<64x128xf32>) -> tensor<64x128xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c16 = constant 16 : index
   %c32 = constant 32 : index
   %zero = constant 0.0 : f32
 
-  %d0 = memref.dim %large_input, %c0 : tensor<64x128xf32>
-  %d1 = memref.dim %large_input, %c1 : tensor<64x128xf32>
+  %d0 = tensor.dim %large_input, %c0 : tensor<64x128xf32>
+  %d1 = tensor.dim %large_input, %c1 : tensor<64x128xf32>
 
   %pad = linalg.pad_tensor %small_input low[4, 60] high[2, 67] {
   ^bb0(%arg0: index, %arg1: index):
     linalg.yield %zero : f32
   } : tensor<58x1xf32> to tensor<64x128xf32>
 
   %fill = linalg.fill(%zero, %large_input) : f32, tensor<64x128xf32> -> tensor<64x128xf32>
 

mlir/test/Dialect/Linalg/tile-tensors.mlir

 // TLOOP-SAME: %[[ARG_2:.*]]: [[TY]]) -> [[TY]] {
 
 // TLOOP-DAG: %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG: %[[C1:.*]] = constant 1 : index
 // TLOOP-DAG: %[[C2:.*]] = constant 2 : index
 // TLOOP-DAG: %[[C3:.*]] = constant 3 : index
 // TLOOP-DAG: %[[C4:.*]] = constant 4 : index
 
-// TLOOP: %[[ARG_0_X:.*]] = memref.dim %[[ARG_0]], %[[C0]] : [[TY]]
-// TLOOP: %[[ARG_0_Y:.*]] = memref.dim %[[ARG_0]], %[[C1]] : [[TY]]
-// TLOOP: %[[ARG_1_Y:.*]] = memref.dim %[[ARG_1]], %[[C1]] : [[TY]]
+// TLOOP: %[[ARG_0_X:.*]] = tensor.dim %[[ARG_0]], %[[C0]] : [[TY]]
+// TLOOP: %[[ARG_0_Y:.*]] = tensor.dim %[[ARG_0]], %[[C1]] : [[TY]]
+// TLOOP: %[[ARG_1_Y:.*]] = tensor.dim %[[ARG_1]], %[[C1]] : [[TY]]
 
 // TLOOP: %{{.*}} = linalg.tiled_loop (%[[I:.*]], %[[J:.*]], %[[K:.*]]) =
 // TLOOP-SAME: (%[[C0]], %[[C0]], %[[C0]])
 // TLOOP-SAME: to (%[[ARG_0_X]], %[[ARG_1_Y]], %[[ARG_0_Y]])
 // TLOOP-SAME: step (%[[C2]], %[[C3]], %[[C4]])
 // TLOOP-SAME: ins (%[[A0:.*]] = %[[ARG_0]]: [[TY]], %[[A1:.*]] = %[[ARG_1]]: [[TY]])
 // TLOOP-SAME: outs (%[[A2:.*]] = %[[ARG_2]]: [[TY]])
 // TLOOP-SAME: iterators["parallel", "parallel", "reduction"]

 
 // -----
 
 func @generic_op_tensors(
   %arg0 : tensor<?x?x?xf32>, %arg1 : tensor<?x?x?xf32>) -> tensor<?x?x?xf32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
-  %0 = memref.dim %arg0, %c0 : tensor<?x?x?xf32>
-  %1 = memref.dim %arg0, %c1 : tensor<?x?x?xf32>
-  %2 = memref.dim %arg0, %c2 : tensor<?x?x?xf32>
+  %0 = tensor.dim %arg0, %c0 : tensor<?x?x?xf32>
+  %1 = tensor.dim %arg0, %c1 : tensor<?x?x?xf32>
+  %2 = tensor.dim %arg0, %c2 : tensor<?x?x?xf32>
   %3 = linalg.init_tensor [%0, %1, %2] : tensor<?x?x?xf32>
   %4 = linalg.generic
     {indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d1, d2)>,
                       affine_map<(d0, d1, d2) -> (d0, d2, d1)>,
                       affine_map<(d0, d1, d2) -> (d2, d1, d0)>],
      iterator_types = ["parallel", "parallel", "parallel"]}
     ins(%arg0, %arg1 : tensor<?x?x?xf32>, tensor<?x?x?xf32>)
     outs(%3 : tensor<?x?x?xf32>) {

 
 // TLOOP-DAG: %[[C0:.*]] = constant 0 : index
 // TLOOP-DAG: %[[C1:.*]] = constant 1 : index
 // TLOOP-DAG: %[[C2:.*]] = constant 2 : index
 // TLOOP-DAG: %[[C3:.*]] = constant 3 : index
 // TLOOP-DAG: %[[C4:.*]] = constant 4 : index
 
 // TLOOP:     %[[INIT:.*]] = linalg.init_tensor
-// TLOOP:     %[[ARG_0_X:.*]] = memref.dim %[[ARG_0]], %[[C0]] : [[TY]]
-// TLOOP:     %[[ARG_0_Y:.*]] = memref.dim %[[ARG_0]], %[[C1]] : [[TY]]
-// TLOOP:     %[[ARG_0_Z:.*]] = memref.dim %[[ARG_0]], %[[C2]] : [[TY]]
+// TLOOP:     %[[ARG_0_X:.*]] = tensor.dim %[[ARG_0]], %[[C0]] : [[TY]]
+// TLOOP:     %[[ARG_0_Y:.*]] = tensor.dim %[[ARG_0]], %[[C1]] : [[TY]]
+// TLOOP:     %[[ARG_0_Z:.*]] = tensor.dim %[[ARG_0]], %[[C2]] : [[TY]]
 
 // TLOOP:     %{{.*}} = linalg.tiled_loop (%{{.*}}, %{{.*}}, %{{.*}}) =
 // TLOOP-SAME: (%[[C0]], %[[C0]], %[[C0]])
 // TLOOP-SAME: to (%[[ARG_0_X]], %[[ARG_0_Y]], %[[ARG_0_Z]])
 // TLOOP-SAME: step (%[[C2]], %[[C3]], %[[C4]])
 // TLOOP-SAME: ins (%{{.*}} = %[[ARG_0]]: [[TY]], %{{.*}} = %[[ARG_1]]: [[TY]])
 // TLOOP-SAME: outs (%{{.*}} = %[[INIT]]: [[TY]])
 // TLOOP-SAME: distribution["block_x", "block_y", "none"] {

mlir/test/Dialect/Linalg/vectorization.mlir

 //  CHECK-SAME:                          %[[SRC:.*]]: tensor<1x2x2x?xf32>, %[[LOW:.*]]: index, %[[HIGH:.*]]: index
 //   CHECK-NOT:   linalg.pad_tensor
 //   CHECK-DAG:   %[[C2:.*]] = constant 2 : index
 //   CHECK-DAG:   %[[C3:.*]] = constant 3 : index
 //   CHECK-DAG:   %[[C5:.*]] = constant 5 : index
 //       CHECK:   %[[V0:.*]] = addi %[[LOW]], %[[C2]] : index
 //       CHECK:   %[[V1:.*]] = addi %[[V0]], %[[C3]] : index
 //       CHECK:   %[[V2:.*]] = addi %[[HIGH]], %[[C5]] : index
-//       CHECK:   %[[DIM3:.*]] = memref.dim %[[SRC]], %[[C3]] : tensor<1x2x2x?xf32>
+//       CHECK:   %[[DIM3:.*]] = tensor.dim %[[SRC]], %[[C3]] : tensor<1x2x2x?xf32>
 //       CHECK:   %[[V4:.*]] = addi %[[DIM3]], %[[C3]] : index
 //       CHECK:   %[[V5:.*]] = addi %[[V4]], %[[C2]] : index
 //       CHECK:   %[[INIT:.*]] = linalg.init_tensor [6, %[[V1]], %[[V2]], %[[V5]]] : tensor<6x?x?x?xf32>
 //       CHECK:   %[[FILL:.*]] = linalg.fill(%{{.*}}, %[[INIT]]) : f32, tensor<6x?x?x?xf32> -> tensor<6x?x?x?xf32>
-//       CHECK:   %[[SRCDIM:.*]] = memref.dim %[[SRC]], %[[C3]] : tensor<1x2x2x?xf32>
+//       CHECK:   %[[SRCDIM:.*]] = tensor.dim %[[SRC]], %[[C3]] : tensor<1x2x2x?xf32>
 //       CHECK:   %[[RESULT:.*]] = tensor.insert_slice %[[SRC]] into %[[FILL]][2, %[[LOW]], 3, 3] [1, 2, 2, %[[SRCDIM]]] [1, 1, 1, 1] : tensor<1x2x2x?xf32> into tensor<6x?x?x?xf32>
 //       CHECK:   return %[[RESULT]]
 func @pad_static_dynamic(%arg0: tensor<1x2x2x?xf32>, %low: index, %high: index,
                   %pad_value: f32) -> tensor<6x?x?x?xf32> {
   %0 = linalg.pad_tensor %arg0 low[2, %low, 3, 3] high[3, 3, %high, 2] {
     ^bb0(%arg1: index, %arg2: index, %arg3: index, %arg4: index):
       linalg.yield %pad_value : f32
     } : tensor<1x2x2x?xf32> to tensor<6x?x?x?xf32>

mlir/test/Dialect/MemRef/canonicalize.mlir

func @load_from_buffer_cast(%arg0: index, %arg1: index, %arg2: tensor<?x?xf32>) -> f32 {
   %0 = memref.buffer_cast %arg2 : memref<?x?xf32>
   %1 = memref.load %0[%arg0, %arg1] : memref<?x?xf32>
   return %1 : f32
 }
 
 // -----
 
 
-// Test case: Basic folding of memref.dim(memref.tensor_load(m)) -> memref.dim(m).
+// Test case: Basic folding of tensor.dim(memref.tensor_load(m)) -> memref.dim(m).
 // CHECK-LABEL: func @dim_of_tensor_load(
 //  CHECK-SAME:     %[[MEMREF:[0-9a-z]*]]: memref<?xf32>
 //       CHECK:   %[[C0:.*]] = constant 0
 //       CHECK:   %[[D:.*]] = memref.dim %[[MEMREF]], %[[C0]]
 //       CHECK:   return %[[D]] : index
 func @dim_of_tensor_load(%arg0: memref<?xf32>) -> index {
   %c0 = constant 0 : index
   %0 = memref.tensor_load %arg0 : memref<?xf32>
-  %1 = memref.dim %0, %c0 : tensor<?xf32>
-  return %1 : index
-}
-
-// -----
-
-// Test case: Folding of memref.dim(tensor.generate %idx) -> %idx
-// CHECK-LABEL: func @dim_of_tensor.generate(
-//  CHECK-SAME:     %[[IDX0:[0-9a-z]+]]: index, %[[IDX1:[0-9a-z]+]]: index
-//   CHECK-NOT:   memref.dim
-//       CHECK:   return %[[IDX1]] : index
-func @dim_of_tensor.generate(%arg0: index, %arg1: index) -> index {
-  %c3 = constant 3 : index
-  %0 = tensor.generate %arg0, %arg1 {
-  ^bb0(%arg2: index, %arg3: index, %arg4: index, %arg5: index, %arg6: index):
-    tensor.yield %c3 : index
-  } : tensor<2x?x4x?x5xindex>
-  %1 = memref.dim %0, %c3 : tensor<2x?x4x?x5xindex>
+  %1 = tensor.dim %0, %c0 : tensor<?xf32>
   return %1 : index
 }
 
 // -----
 
 // Test case: Folding of memref.dim(memref.alloca(%size), %idx) -> %size
 // CHECK-LABEL: func @dim_of_alloca(
 //  CHECK-SAME:     %[[SIZE:[0-9a-z]+]]: index
func @dim_of_memref_reshape_i32(%arg0: memref<*xf32>, %arg1: memref<?xi32>)
   %0 = memref.reshape %arg0(%arg1)
       : (memref<*xf32>, memref<?xi32>) -> memref<*xf32>
   %1 = memref.dim %0, %c3 : memref<*xf32>
   return %1 : index
 }
 
 // -----
 
-// Test case: Folding memref.dim(tensor.cast %0, %idx) -> memref.dim %0, %idx
-// CHECK-LABEL: func @fold_dim_of_tensor.cast
-//  CHECK-SAME:   %[[ARG0:.[a-z0-9A-Z_]+]]: tensor<4x?xf32>
-//   CHECK-DAG:   %[[C1:.+]] = constant 1 : index
-//   CHECK-DAG:   %[[C4:.+]] = constant 4 : index
-//       CHECK:   %[[T0:.+]] = memref.dim %[[ARG0]], %[[C1]]
-//  CHECK-NEXT:   return %[[C4]], %[[T0]]
-func @fold_dim_of_tensor.cast(%arg0 : tensor<4x?xf32>) -> (index, index) {
-  %c0 = constant 0 : index
-  %c1 = constant 1 : index
-  %0 = tensor.cast %arg0 : tensor<4x?xf32> to tensor<?x?xf32>
-  %1 = memref.dim %0, %c0 : tensor<?x?xf32>
-  %2 = memref.dim %0, %c1 : tensor<?x?xf32>
-  return %1, %2: index, index
-}
-
-// -----
-
 // CHECK-LABEL: func @tensor_cast_to_memref
 //  CHECK-SAME:   %[[ARG0:.+]]: tensor<4x6x16x32xi8>
 //       CHECK:   %[[M:.+]] = memref.buffer_cast %[[ARG0]] : memref<4x6x16x32xi8>
 //       CHECK:   %[[M1:.+]] = memref.cast %[[M]] : memref<4x6x16x32xi8> to memref<?x?x16x32xi8>
 //       CHECK:   return %[[M1]] : memref<?x?x16x32xi8>
 func @tensor_cast_to_memref(%arg0 : tensor<4x6x16x32xi8>) ->
   memref<?x?x16x32xi8> {
   %0 = tensor.cast %arg0 : tensor<4x6x16x32xi8> to tensor<?x?x16x32xi8>

mlir/test/Dialect/SparseTensor/conversion.mlir

 
 // CHECK-LABEL: func @sparse_dim(
 //  CHECK-SAME: %[[A:.*]]: !llvm.ptr<i8>)
 //       CHECK: %[[C:.*]] = constant 0 : index
 //       CHECK: %[[D:.*]] = call @sparseDimSize(%[[A]], %[[C]])
 //       CHECK: return %[[D]] : index
 func @sparse_dim(%arg0: tensor<?xf64, #SparseVector>) -> index {
   %c = constant 0 : index
-  %0 = memref.dim %arg0, %c : tensor<?xf64, #SparseVector>
+  %0 = tensor.dim %arg0, %c : tensor<?xf64, #SparseVector>
   return %0 : index
 }
 
 // CHECK-LABEL: func @sparse_new1d(
 //  CHECK-SAME: %[[A:.*]]: !llvm.ptr<i8>) -> !llvm.ptr<i8>
 //       CHECK: %[[D:.*]] = constant dense<1> : tensor<1xi8>
 //       CHECK: %[[C:.*]] = tensor.cast %[[D]] : tensor<1xi8> to tensor<?xi8>
 //       CHECK: %[[P:.*]] = constant dense<0> : tensor<1xi64>

mlir/test/Dialect/SparseTensor/sparse_1d.mlir

 // CHECK:           %[[VAL_8:.*]] = memref.buffer_cast %[[VAL_0]] : memref<?xf64>
 // CHECK:           %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_5]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_5]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf64>
 // CHECK:           %[[VAL_12:.*]] = memref.buffer_cast %[[VAL_2]] : memref<?xf64>
 // CHECK:           %[[VAL_13:.*]] = sparse_tensor.pointers %[[VAL_3]], %[[VAL_5]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = sparse_tensor.indices %[[VAL_3]], %[[VAL_5]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = sparse_tensor.values %[[VAL_3]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf64>
-// CHECK:           %[[VAL_16:.*]] = memref.dim %[[VAL_4]], %[[VAL_5]] : tensor<?xf64>
+// CHECK:           %[[VAL_16:.*]] = tensor.dim %[[VAL_4]], %[[VAL_5]] : tensor<?xf64>
 // CHECK:           %[[VAL_17:.*]] = memref.buffer_cast %[[VAL_4]] : memref<?xf64>
 // CHECK:           %[[VAL_18:.*]] = memref.alloc(%[[VAL_16]]) : memref<?xf64>
 // CHECK:           linalg.copy(%[[VAL_17]], %[[VAL_18]]) : memref<?xf64>, memref<?xf64>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_20:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_21:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_22:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_23:.*]]:3 = scf.while (%[[VAL_24:.*]] = %[[VAL_19]], %[[VAL_25:.*]] = %[[VAL_21]], %[[VAL_26:.*]] = %[[VAL_5]]) : (index, index, index) -> (index, index, index) {

mlir/test/Dialect/SparseTensor/sparse_2d.mlir

 // CHECK-SAME:                %[[VAL_0:.*]]: tensor<?x?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:                %[[VAL_1:.*]]: tensor<?x?xf64>) -> tensor<?x?xf64> {
 // CHECK:           %[[VAL_2:.*]] = constant 2.000000e+00 : f64
 // CHECK:           %[[VAL_3:.*]] = constant 0 : index
 // CHECK:           %[[VAL_4:.*]] = constant 1 : index
 // CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<?x?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<?x?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<?x?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf64>
-// CHECK:           %[[VAL_8:.*]] = memref.dim %[[VAL_1]], %[[VAL_3]] : tensor<?x?xf64>
-// CHECK:           %[[VAL_9:.*]] = memref.dim %[[VAL_1]], %[[VAL_4]] : tensor<?x?xf64>
+// CHECK:           %[[VAL_8:.*]] = tensor.dim %[[VAL_1]], %[[VAL_3]] : tensor<?x?xf64>
+// CHECK:           %[[VAL_9:.*]] = tensor.dim %[[VAL_1]], %[[VAL_4]] : tensor<?x?xf64>
 // CHECK:           %[[VAL_10:.*]] = memref.buffer_cast %[[VAL_1]] : memref<?x?xf64>
 // CHECK:           %[[VAL_11:.*]] = memref.alloc(%[[VAL_8]], %[[VAL_9]]) : memref<?x?xf64>
 // CHECK:           linalg.copy(%[[VAL_10]], %[[VAL_11]]) : memref<?x?xf64>, memref<?x?xf64>
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_3]] to %[[VAL_8]] step %[[VAL_4]] {
 // CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_12]]] : memref<?xindex>
 // CHECK:             %[[VAL_14:.*]] = addi %[[VAL_12]], %[[VAL_4]] : index
 // CHECK:             %[[VAL_15:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_14]]] : memref<?xindex>
 // CHECK:             scf.for %[[VAL_16:.*]] = %[[VAL_13]] to %[[VAL_15]] step %[[VAL_4]] {

 // CHECK:           %[[VAL_4:.*]] = constant 0 : index
 // CHECK:           %[[VAL_5:.*]] = constant 1 : index
 // CHECK:           %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_5]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_5]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK:           %[[VAL_11:.*]] = memref.buffer_cast %[[VAL_1]] : memref<?x?xf32>
-// CHECK:           %[[VAL_12:.*]] = memref.dim %[[VAL_2]], %[[VAL_4]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_12:.*]] = tensor.dim %[[VAL_2]], %[[VAL_4]] : tensor<?x?xf32>
 // CHECK:           %[[VAL_13:.*]] = memref.buffer_cast %[[VAL_2]] : memref<?x?xf32>
-// CHECK:           %[[VAL_14:.*]] = memref.dim %[[VAL_3]], %[[VAL_4]] : tensor<?x?xf32>
-// CHECK:           %[[VAL_15:.*]] = memref.dim %[[VAL_3]], %[[VAL_5]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_14:.*]] = tensor.dim %[[VAL_3]], %[[VAL_4]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_15:.*]] = tensor.dim %[[VAL_3]], %[[VAL_5]] : tensor<?x?xf32>
 // CHECK:           %[[VAL_16:.*]] = memref.buffer_cast %[[VAL_3]] : memref<?x?xf32>
 // CHECK:           %[[VAL_17:.*]] = memref.alloc(%[[VAL_14]], %[[VAL_15]]) : memref<?x?xf32>
 // CHECK:           linalg.copy(%[[VAL_16]], %[[VAL_17]]) : memref<?x?xf32>, memref<?x?xf32>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_20:.*]] = %[[VAL_18]] to %[[VAL_19]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_21:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_20]]] : memref<?xindex>
 // CHECK:             scf.for %[[VAL_22:.*]] = %[[VAL_4]] to %[[VAL_12]] step %[[VAL_5]] {

 // CHECK:           %[[VAL_14:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK:           %[[VAL_17:.*]] = sparse_tensor.pointers %[[VAL_2]], %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = sparse_tensor.indices %[[VAL_2]], %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = sparse_tensor.values %[[VAL_2]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK:           %[[VAL_20:.*]] = memref.buffer_cast %[[VAL_3]] : memref<?xf32>
 // CHECK:           %[[VAL_21:.*]] = memref.buffer_cast %[[VAL_4]] : memref<f32>
-// CHECK:           %[[VAL_22:.*]] = memref.dim %[[VAL_5]], %[[VAL_6]] : tensor<?xf32>
+// CHECK:           %[[VAL_22:.*]] = tensor.dim %[[VAL_5]], %[[VAL_6]] : tensor<?xf32>
 // CHECK:           %[[VAL_23:.*]] = memref.buffer_cast %[[VAL_5]] : memref<?xf32>
 // CHECK:           %[[VAL_24:.*]] = memref.alloc(%[[VAL_22]]) : memref<?xf32>
 // CHECK:           linalg.copy(%[[VAL_23]], %[[VAL_24]]) : memref<?xf32>, memref<?xf32>
 // CHECK:           %[[VAL_25:.*]] = memref.load %[[VAL_21]][] : memref<f32>
 // CHECK:           %[[VAL_26:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_27:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:           %[[VAL_28:.*]]:2 = scf.while (%[[VAL_29:.*]] = %[[VAL_26]], %[[VAL_30:.*]] = %[[VAL_6]]) : (index, index) -> (index, index) {
 // CHECK:             %[[VAL_31:.*]] = cmpi ult, %[[VAL_29]], %[[VAL_27]] : index

mlir/test/Dialect/SparseTensor/sparse_3d.mlir

 // CHECK-SAME:                    %[[VAL_2:.*2]]: tensor<?x?xf32>,
 // CHECK-SAME:                    %[[VAL_3:.*3]]: tensor<?x?xf32>) -> tensor<?x?xf32> {
 // CHECK:           %[[VAL_4:.*]] = constant 2 : index
 // CHECK:           %[[VAL_5:.*]] = constant 0 : index
 // CHECK:           %[[VAL_6:.*]] = constant 1 : index
 // CHECK:           %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<?x?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<?x?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK:           %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?x?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK:           %[[VAL_10:.*]] = memref.dim %[[VAL_2]], %[[VAL_5]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_10:.*]] = tensor.dim %[[VAL_2]], %[[VAL_5]] : tensor<?x?xf32>
 // CHECK:           %[[VAL_11:.*]] = memref.buffer_cast %[[VAL_2]] : memref<?x?xf32>
 // CHECK:           %[[VAL_12:.*]] = memref.buffer_cast %[[VAL_3]] : memref<?x?xf32>
-// CHECK:           %[[VAL_13:.*]] = memref.dim %[[VAL_0]], %[[VAL_5]] : tensor<?x?xf32>
-// CHECK:           %[[VAL_14:.*]] = memref.dim %[[VAL_0]], %[[VAL_6]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_13:.*]] = tensor.dim %[[VAL_0]], %[[VAL_5]] : tensor<?x?xf32>
+// CHECK:           %[[VAL_14:.*]] = tensor.dim %[[VAL_0]], %[[VAL_6]] : tensor<?x?xf32>
 // CHECK:           %[[VAL_15:.*]] = memref.buffer_cast %[[VAL_0]] : memref<?x?xf32>
 // CHECK:           %[[VAL_16:.*]] = memref.alloc(%[[VAL_13]], %[[VAL_14]]) : memref<?x?xf32>
 // CHECK:           linalg.copy(%[[VAL_15]], %[[VAL_16]]) : memref<?x?xf32>, memref<?x?xf32>
 // CHECK:           scf.for %[[VAL_17:.*]] = %[[VAL_5]] to %[[VAL_13]] step %[[VAL_6]] {
 // CHECK:             scf.for %[[VAL_18:.*]] = %[[VAL_5]] to %[[VAL_10]] step %[[VAL_6]] {
 // CHECK:               %[[VAL_19:.*]] = muli %[[VAL_10]], %[[VAL_17]] : index
 // CHECK:               %[[VAL_20:.*]] = addi %[[VAL_19]], %[[VAL_18]] : index
 // CHECK:               %[[VAL_21:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_20]]] : memref<?xindex>

 
 // CHECK-LABEL:   func @sum_reduction_inv(
 // CHECK-SAME:                            %[[VAL_0:.*]]: tensor<?x?x?xf32>,
 // CHECK-SAME:                            %[[VAL_1:.*]]: tensor<?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:                            %[[VAL_2:.*]]: tensor<f32>) -> tensor<f32> {
 // CHECK:           %[[VAL_3:.*]] = constant 2 : index
 // CHECK:           %[[VAL_4:.*]] = constant 0 : index
 // CHECK:           %[[VAL_5:.*]] = constant 1 : index
-// CHECK:           %[[VAL_6:.*]] = memref.dim %[[VAL_0]], %[[VAL_5]] : tensor<?x?x?xf32>
-// CHECK:           %[[VAL_7:.*]] = memref.dim %[[VAL_0]], %[[VAL_3]] : tensor<?x?x?xf32>
+// CHECK:           %[[VAL_6:.*]] = tensor.dim %[[VAL_0]], %[[VAL_5]] : tensor<?x?x?xf32>
+// CHECK:           %[[VAL_7:.*]] = tensor.dim %[[VAL_0]], %[[VAL_3]] : tensor<?x?x?xf32>
 // CHECK:           %[[VAL_8:.*]] = memref.buffer_cast %[[VAL_0]] : memref<?x?x?xf32>
-// CHECK:           %[[VAL_9:.*]] = memref.dim %[[VAL_1]], %[[VAL_4]] : tensor<?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>
+// CHECK:           %[[VAL_9:.*]] = tensor.dim %[[VAL_1]], %[[VAL_4]] : tensor<?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>
 // CHECK:           %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK:           %[[VAL_11:.*]] = memref.buffer_cast %[[VAL_2]] : memref<f32>
 // CHECK:           %[[VAL_12:.*]] = memref.alloc() : memref<f32>
 // CHECK:           linalg.copy(%[[VAL_11]], %[[VAL_12]]) : memref<f32>, memref<f32>
 // CHECK:           scf.for %[[VAL_13:.*]] = %[[VAL_4]] to %[[VAL_9]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_13]]] : memref<?xf32>
 // CHECK:             scf.for %[[VAL_15:.*]] = %[[VAL_4]] to %[[VAL_6]] step %[[VAL_5]] {
 // CHECK:               %[[VAL_16:.*]] = memref.load %[[VAL_12]][] : memref<f32>

mlir/test/Dialect/Standard/bufferize.mlir

 // RUN: mlir-opt %s -std-bufferize | FileCheck %s
 
 // CHECK-LABEL:   func @dim(
 // CHECK-SAME:              %[[TENSOR:.*]]: tensor<f32>,
 // CHECK-SAME:              %[[INDEX:.*]]: index) -> index {
 // CHECK:           %[[MEMREF:.*]] = memref.buffer_cast %[[TENSOR]] : memref<f32>
 // CHECK:           %[[EXTENT:.*]] = memref.dim %[[MEMREF]], %[[INDEX]] : memref<f32>
 // CHECK:           return %[[EXTENT]] : index
 func @dim(%arg0: tensor<f32>, %arg1: index) -> index {
-  %0 = memref.dim %arg0, %arg1 : tensor<f32>
+  %0 = tensor.dim %arg0, %arg1 : tensor<f32>
   return %0 : index
 }
 
 // CHECK-LABEL:   func @select(
 // CHECK-SAME:                 %[[PRED:.*]]: i1,
 // CHECK-SAME:                 %[[TRUE_VAL:.*]]: tensor<f32>,
 // CHECK-SAME:                 %[[FALSE_VAL:.*]]: tensor<f32>) -> tensor<f32> {
 // CHECK:           %[[TRUE_VAL_MEMREF:.*]] = memref.buffer_cast %[[TRUE_VAL]] : memref<f32>

mlir/test/Dialect/Tensor/bufferize.mlir

 
 // CHECK-LABEL:   func @tensor.generate(
 // CHECK-SAME:                                       %[[ARG:.*]]: tensor<*xf32>,
 // CHECK-SAME:                                       %[[DYNAMIC_EXTENT:.*]]: index) -> tensor<?xindex> {
 // CHECK:           %[[MEMREF:.*]] = memref.alloc(%[[DYNAMIC_EXTENT]]) : memref<?xindex>
 // CHECK:           %[[C0:.*]] = constant 0 : index
 // CHECK:           %[[C1:.*]] = constant 1 : index
 // CHECK:           scf.parallel (%[[I:.*]]) = (%[[C0]]) to (%[[DYNAMIC_EXTENT]]) step (%[[C1]]) {
-// CHECK:             %[[ELEM:.*]] = memref.dim %[[ARG]], %[[I]] : tensor<*xf32>
+// CHECK:             %[[ELEM:.*]] = tensor.dim %[[ARG]], %[[I]] : tensor<*xf32>
 // CHECK:             store %[[ELEM]], %[[MEMREF]][%[[I]]] : memref<?xindex>
 // CHECK:             scf.yield
 // CHECK:           }
 // CHECK:           %[[RET:.*]] = memref.tensor_load %[[MEMREF]] : memref<?xindex>
 // CHECK:           return %[[RET]] : tensor<?xindex>
 // CHECK:         }
 func @tensor.generate(%arg: tensor<*xf32>, %dynamic_extent: index) -> tensor<?xindex> {
   %result = tensor.generate %dynamic_extent {
   ^bb0(%i : index):
-    %elem = memref.dim %arg, %i : tensor<*xf32>
+    %elem = tensor.dim %arg, %i : tensor<*xf32>
     tensor.yield %elem : index
   } : tensor<?xindex>
   return %result : tensor<?xindex>
 }
 
 // Additional test that checks the logic for intermixed static and dynamic
 // extents.
 //

mlir/test/Dialect/Tensor/canonicalize.mlir

 }
 
 // -----
 
 // CHECK-LABEL: func @extract_from_tensor.generate
 // CHECK-SAME: %[[IDX:.*]]: index, %[[TENSOR:.*]]: tensor<*xf32>
 func @extract_from_tensor.generate(%idx: index, %tensor: tensor<*xf32>) -> index {
   %size = rank %tensor : tensor<*xf32>
-  // CHECK-NEXT: %[[RES:.*]] = memref.dim %[[TENSOR]], %[[IDX]]
+  // CHECK-NEXT: %[[RES:.*]] = tensor.dim %[[TENSOR]], %[[IDX]]
   %0 = tensor.generate %size {
     ^bb0(%arg0: index):
-    %1 = memref.dim %tensor, %arg0 : tensor<*xf32>
+    %1 = tensor.dim %tensor, %arg0 : tensor<*xf32>
     tensor.yield %1 : index
   } : tensor<?xindex>
   %1 = tensor.extract %0[%idx] : tensor<?xindex>
   // CHECK-NEXT: return %[[RES]]
   return %1 : index
 }
 
 // -----
 
 // CHECK-LABEL: func @extract_from_tensor.generate_2d
 // CHECK-SAME: %[[IDX0:.*]]: index, %[[IDX1:.*]]: index, %[[TENSOR:.*]]: tensor<*xf32>
 func @extract_from_tensor.generate_2d(%idx0: index, %idx1: index, %tensor: tensor<*xf32>) -> index {
   %size = rank %tensor : tensor<*xf32>
-  // CHECK-NEXT: %[[DIM0:.*]] = memref.dim %[[TENSOR]], %[[IDX0]]
-  // CHECK-NEXT: %[[DIM1:.*]] = memref.dim %[[TENSOR]], %[[IDX1]]
+  // CHECK-NEXT: %[[DIM0:.*]] = tensor.dim %[[TENSOR]], %[[IDX0]]
+  // CHECK-NEXT: %[[DIM1:.*]] = tensor.dim %[[TENSOR]], %[[IDX1]]
   // CHECK-NEXT: %[[RES:.*]] = addi %[[DIM0]], %[[DIM1]]
   %0 = tensor.generate %size, %size {
     ^bb0(%arg0: index, %arg1: index):
-    %1 = memref.dim %tensor, %arg0 : tensor<*xf32>
-    %2 = memref.dim %tensor, %arg1 : tensor<*xf32>
+    %1 = tensor.dim %tensor, %arg0 : tensor<*xf32>
+    %2 = tensor.dim %tensor, %arg1 : tensor<*xf32>
     %3 = addi %1, %2 : index
     tensor.yield %3 : index
   } : tensor<?x?xindex>
   %4 = tensor.extract %0[%idx0, %idx1] : tensor<?x?xindex>
   // CHECK-NEXT: return %[[RES]]
   return %4 : index
 }
 
 // -----
 
 // CHECK-LABEL: func @extract_from_tensor.generate_sideeffects
 // CHECK-SAME: %[[IDX:.*]]: index
 func @extract_from_tensor.generate_sideeffects(%idx: index, %tensor: tensor<*xf32>, %mem: memref<?xindex>) -> index {
   %size = rank %tensor : tensor<*xf32>
   // CHECK: %[[DTENSOR:.*]] = tensor.generate
   %0 = tensor.generate %size {
     ^bb0(%arg0: index):
-    %1 = memref.dim %tensor, %arg0 : tensor<*xf32>
+    %1 = tensor.dim %tensor, %arg0 : tensor<*xf32>
     memref.store %1, %mem[%arg0] : memref<?xindex>
     tensor.yield %1 : index
   } : tensor<?xindex>
   // CHECK: %[[RES:.*]] = tensor.extract %[[DTENSOR]][%[[IDX]]]
   %1 = tensor.extract %0[%idx] : tensor<?xindex>
   // CHECK-NEXT: return %[[RES]]
   return %1 : index
 }

 // -----
 
 func @insert_slice_propagate_dest_cast(%arg0 : tensor<2x?xi32>, %arg1 : tensor<i32>,
     %arg2 : index, %arg3 : index) -> tensor<?x?xi32> {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %c8 = constant 8 : index
-  %0 = memref.dim %arg0, %c1 : tensor<2x?xi32>
+  %0 = tensor.dim %arg0, %c1 : tensor<2x?xi32>
   %1 = tensor.extract %arg1[] : tensor<i32>
   %2 = tensor.generate %arg2, %c8 {
   ^bb0(%arg4: index, %arg5: index):
     tensor.yield %1 : i32
   } : tensor<?x?xi32>
   %3 = tensor.insert_slice %arg0 into %2[%c0, %arg3] [%c2, %0] [%c1, %c1] : tensor<2x?xi32> into tensor<?x?xi32>
   return %3 : tensor<?x?xi32>
 }

 }
 // CHECK-LABEL: func @insert_slice_output_dest_canonicalize
 //  CHECK-SAME:   %[[ARG0:[a-zA-z0-9_]+]]: tensor<2x3xi32>
 //  CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: tensor<i32>
 //       CHECK:   %[[PAD:.+]] = tensor.extract %[[ARG1]]
 //       CHECK:   %[[GENERATE:.+]] = tensor.generate
 //       CHECK:   %[[RESULT:.+]] = tensor.insert_slice %[[ARG0]] into %[[GENERATE]]
 //       CHECK:   return %[[RESULT]]
+
+// -----
+
+// Test case: Folding of tensor.dim(tensor.generate %idx) -> %idx
+// CHECK-LABEL: func @dim_of_tensor.generate(
+//  CHECK-SAME:     %[[IDX0:[0-9a-z]+]]: index, %[[IDX1:[0-9a-z]+]]: index
+//   CHECK-NOT:   tensor.dim
+//       CHECK:   return %[[IDX1]] : index
+func @dim_of_tensor.generate(%arg0: index, %arg1: index) -> index {
+  %c3 = constant 3 : index
+  %0 = tensor.generate %arg0, %arg1 {
+  ^bb0(%arg2: index, %arg3: index, %arg4: index, %arg5: index, %arg6: index):
+    tensor.yield %c3 : index
+  } : tensor<2x?x4x?x5xindex>
+  %1 = tensor.dim %0, %c3 : tensor<2x?x4x?x5xindex>
+  return %1 : index
+}
+
+// -----
+
+// Test case: Folding tensor.dim(tensor.cast %0, %idx) -> tensor.dim %0, %idx
+// CHECK-LABEL: func @fold_dim_of_tensor.cast
+//  CHECK-SAME:   %[[ARG0:.[a-z0-9A-Z_]+]]: tensor<4x?xf32>
+//   CHECK-DAG:   %[[C1:.+]] = constant 1 : index
+//   CHECK-DAG:   %[[C4:.+]] = constant 4 : index
+//       CHECK:   %[[T0:.+]] = tensor.dim %[[ARG0]], %[[C1]]
+//  CHECK-NEXT:   return %[[C4]], %[[T0]]
+func @fold_dim_of_tensor.cast(%arg0 : tensor<4x?xf32>) -> (index, index) {
+  %c0 = constant 0 : index
+  %c1 = constant 1 : index
+  %0 = tensor.cast %arg0 : tensor<4x?xf32> to tensor<?x?xf32>
+  %1 = tensor.dim %0, %c0 : tensor<?x?xf32>
+  %2 = tensor.dim %0, %c1 : tensor<?x?xf32>
+  return %1, %2: index, index
+}

mlir/test/IR/core-ops.mlir

 // CHECK-LABEL: func @func_with_ops
 // CHECK-SAME: %[[ARG:.*]]: f32
 func @func_with_ops(f32) {
 ^bb0(%a : f32):
   // CHECK: %[[T:.*]] = "getTensor"() : () -> tensor<4x4x?xf32>
   %t = "getTensor"() : () -> tensor<4x4x?xf32>
 
   // CHECK: %[[C2:.*]] = constant 2 : index
-  // CHECK-NEXT: %{{.*}} = memref.dim %[[T]], %[[C2]] : tensor<4x4x?xf32>
+  // CHECK-NEXT: %{{.*}} = tensor.dim %[[T]], %[[C2]] : tensor<4x4x?xf32>
   %c2 = constant 2 : index
-  %t2 = "memref.dim"(%t, %c2) : (tensor<4x4x?xf32>, index) -> index
+  %t2 = "tensor.dim"(%t, %c2) : (tensor<4x4x?xf32>, index) -> index
 
   // CHECK: %{{.*}} = addf %[[ARG]], %[[ARG]] : f32
   %x = "std.addf"(%a, %a) : (f32,f32) -> (f32)
 
   // CHECK: return
   return
 }
 
 // CHECK-LABEL: func @standard_instrs(%arg0: tensor<4x4x?xf32>, %arg1: f32, %arg2: i32, %arg3: index, %arg4: i64, %arg5: f16) {
 func @standard_instrs(tensor<4x4x?xf32>, f32, i32, index, i64, f16) {
 ^bb42(%t: tensor<4x4x?xf32>, %f: f32, %i: i32, %idx : index, %j: i64, %half: f16):
   // CHECK: %[[C2:.*]] = constant 2 : index
-  // CHECK: %[[A2:.*]] = memref.dim %arg0, %[[C2]] : tensor<4x4x?xf32>
+  // CHECK: %[[A2:.*]] = tensor.dim %arg0, %[[C2]] : tensor<4x4x?xf32>
   %c2 = constant 2 : index
-  %a2 = memref.dim %t, %c2 : tensor<4x4x?xf32>
+  %a2 = tensor.dim %t, %c2 : tensor<4x4x?xf32>
 
   // CHECK: %[[F2:.*]] = addf %arg1, %arg1 : f32
   %f2 = "std.addf"(%f, %f) : (f32,f32) -> f32
 
   // CHECK: %[[F3:.*]] = addf %[[F2]], %[[F2]] : f32
   %f3 = addf %f2, %f2 : f32
 
   // CHECK: %[[I2:.*]] = addi %arg2, %arg2 : i32
func @memref_subview(%arg0 : index, %arg1 : index, %arg2 : index) {
 
   return
 }
 
 // CHECK-LABEL: func @test_dimop
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x4x?xf32>
 func @test_dimop(%arg0: tensor<4x4x?xf32>) {
   // CHECK: %[[C2:.*]] = constant 2 : index
-  // CHECK: %{{.*}} = memref.dim %[[ARG]], %[[C2]] : tensor<4x4x?xf32>
+  // CHECK: %{{.*}} = tensor.dim %[[ARG]], %[[C2]] : tensor<4x4x?xf32>
   %c2 = constant 2 : index
-  %0 = memref.dim %arg0, %c2 : tensor<4x4x?xf32>
+  %0 = tensor.dim %arg0, %c2 : tensor<4x4x?xf32>
   // use dim as an index to ensure type correctness
   %1 = affine.apply affine_map<(d0) -> (d0)>(%0)
   return
 }
 
 // CHECK-LABEL: func @test_splat_op
 // CHECK-SAME: [[S:%arg[0-9]+]]: f32
 func @test_splat_op(%s : f32) {

mlir/test/IR/invalid-ops.mlir

 // RUN: mlir-opt -allow-unregistered-dialect %s -split-input-file -verify-diagnostics
 
 func @dim(%arg : tensor<1x?xf32>) {
   %c2 = constant 2 : index
-  memref.dim %arg, %c2 : tensor<1x?xf32> // expected-error {{'memref.dim' op index is out of range}}
+  tensor.dim %arg, %c2 : tensor<1x?xf32> // expected-error {{'tensor.dim' op index is out of range}}
   return
 }
 
 // -----
 
 func @rank(f32) {
 ^bb(%0: f32):
   "std.rank"(%0): (f32)->index // expected-error {{'std.rank' op operand #0 must be any memref or tensor type}}

mlir/test/Interfaces/InferShapedTypeOpInterface/resolve-shaped-type-result-dims.mlir

 // RUN: mlir-opt %s -resolve-shaped-type-result-dims -split-input-file | FileCheck %s
 
 func @result_shape(%arg0 : tensor<2x3x?xf32>, %arg1 : tensor<?x5xf32>)
     -> (index, index, index, index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %0:2 = "test.op_with_result_shape_interface"(%arg0, %arg1)
       : (tensor<2x3x?xf32>, tensor<?x5xf32>) -> (tensor<?x5xf32>, tensor<2x3x?xf32>)
-  %1 = memref.dim %0#0, %c0 : tensor<?x5xf32>
-  %2 = memref.dim %0#0, %c1 : tensor<?x5xf32>
-  %3 = memref.dim %0#1, %c0 : tensor<2x3x?xf32>
-  %4 = memref.dim %0#1, %c1 : tensor<2x3x?xf32>
-  %5 = memref.dim %0#1, %c2 : tensor<2x3x?xf32>
+  %1 = tensor.dim %0#0, %c0 : tensor<?x5xf32>
+  %2 = tensor.dim %0#0, %c1 : tensor<?x5xf32>
+  %3 = tensor.dim %0#1, %c0 : tensor<2x3x?xf32>
+  %4 = tensor.dim %0#1, %c1 : tensor<2x3x?xf32>
+  %5 = tensor.dim %0#1, %c2 : tensor<2x3x?xf32>
   return %1, %2, %3, %4, %5 : index, index, index, index, index
 }
 // CHECK-LABEL: func @result_shape(
 //  CHECK-SAME:   %[[ARG_0:[a-z0-9]*]]: tensor<2x3x?xf32>
 //  CHECK-SAME:   %[[ARG_1:[a-z0-9]*]]: tensor<?x5xf32>)
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //   CHECK-DAG:   %[[C3:.+]] = constant 3 : index
 //   CHECK-DAG:   %[[C5:.+]] = constant 5 : index
-//   CHECK-DAG:   %[[D0:.+]] = memref.dim %[[ARG_1]], %[[C0]]
+//   CHECK-DAG:   %[[D0:.+]] = tensor.dim %[[ARG_1]], %[[C0]]
 //   CHECK-DAG:   %[[S0:.+]] = tensor.from_elements %[[D0]], %[[C5]]
 //   CHECK-DAG:   %[[D0_OUT:.+]] = tensor.extract %[[S0]][%[[C0]]]
-//   CHECK-DAG:   %[[D1:.+]] = memref.dim %[[ARG_0]], %[[C2]]
+//   CHECK-DAG:   %[[D1:.+]] = tensor.dim %[[ARG_0]], %[[C2]]
 //   CHECK-DAG:   %[[S1:.+]] = tensor.from_elements %[[C2]], %[[C3]], %[[D1]]
 //   CHECK-DAG:   %[[D1_OUT:.+]] = tensor.extract %[[S1]][%[[C2]]]
 //       CHECK:   return %[[D0_OUT]], %[[C5]], %[[C2]], %[[C3]], %[[D1_OUT]]
 
 // -----
 
 func @result_shape_per_dim(%arg0 : tensor<2x3x?xf32>, %arg1 : tensor<?x5xf32>)
     -> (index, index, index, index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %0:2 = "test.op_with_result_shape_per_dim_interface"(%arg0, %arg1)
       : (tensor<2x3x?xf32>, tensor<?x5xf32>) -> (tensor<?x5xf32>, tensor<2x3x?xf32>)
-  %1 = memref.dim %0#0, %c0 : tensor<?x5xf32>
-  %2 = memref.dim %0#0, %c1 : tensor<?x5xf32>
-  %3 = memref.dim %0#1, %c0 : tensor<2x3x?xf32>
-  %4 = memref.dim %0#1, %c1 : tensor<2x3x?xf32>
-  %5 = memref.dim %0#1, %c2 : tensor<2x3x?xf32>
+  %1 = tensor.dim %0#0, %c0 : tensor<?x5xf32>
+  %2 = tensor.dim %0#0, %c1 : tensor<?x5xf32>
+  %3 = tensor.dim %0#1, %c0 : tensor<2x3x?xf32>
+  %4 = tensor.dim %0#1, %c1 : tensor<2x3x?xf32>
+  %5 = tensor.dim %0#1, %c2 : tensor<2x3x?xf32>
   return %1, %2, %3, %4, %5 : index, index, index, index, index
 }
 // CHECK-LABEL: func @result_shape_per_dim(
 //  CHECK-SAME:   %[[ARG_0:[a-z0-9]*]]: tensor<2x3x?xf32>
 //  CHECK-SAME:   %[[ARG_1:[a-z0-9]*]]: tensor<?x5xf32>)
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //   CHECK-DAG:   %[[C3:.+]] = constant 3 : index
 //   CHECK-DAG:   %[[C5:.+]] = constant 5 : index
-//   CHECK-DAG:   %[[D0:.+]] = memref.dim %[[ARG_1]], %[[C0]]
-//   CHECK-DAG:   %[[D1:.+]] = memref.dim %[[ARG_0]], %[[C2]]
+//   CHECK-DAG:   %[[D0:.+]] = tensor.dim %[[ARG_1]], %[[C0]]
+//   CHECK-DAG:   %[[D1:.+]] = tensor.dim %[[ARG_0]], %[[C2]]
 //       CHECK:   return %[[D0]], %[[C5]], %[[C2]], %[[C3]], %[[D1]]
 
 // -----
 
 func @result_shape_and_per_dim(%arg0 : tensor<2x3x?xf32>, %arg1 : tensor<?x5xf32>)
     -> (index, index, index, index, index) {
   %c0 = constant 0 : index
   %c1 = constant 1 : index
   %c2 = constant 2 : index
   %0:2 = "test.op_with_result_shape_and_per_dim_interface"(%arg0, %arg1)
       : (tensor<2x3x?xf32>, tensor<?x5xf32>) -> (tensor<?x5xf32>, tensor<2x3x?xf32>)
-  %1 = memref.dim %0#0, %c0 : tensor<?x5xf32>
-  %2 = memref.dim %0#0, %c1 : tensor<?x5xf32>
-  %3 = memref.dim %0#1, %c0 : tensor<2x3x?xf32>
-  %4 = memref.dim %0#1, %c1 : tensor<2x3x?xf32>
-  %5 = memref.dim %0#1, %c2 : tensor<2x3x?xf32>
+  %1 = tensor.dim %0#0, %c0 : tensor<?x5xf32>
+  %2 = tensor.dim %0#0, %c1 : tensor<?x5xf32>
+  %3 = tensor.dim %0#1, %c0 : tensor<2x3x?xf32>
+  %4 = tensor.dim %0#1, %c1 : tensor<2x3x?xf32>
+  %5 = tensor.dim %0#1, %c2 : tensor<2x3x?xf32>
   return %1, %2, %3, %4, %5 : index, index, index, index, index
 }
 // CHECK-LABEL: func @result_shape_and_per_dim(
 //  CHECK-SAME:   %[[ARG_0:[a-z0-9]*]]: tensor<2x3x?xf32>
 //  CHECK-SAME:   %[[ARG_1:[a-z0-9]*]]: tensor<?x5xf32>)
 //   CHECK-DAG:   %[[C0:.+]] = constant 0 : index
 //   CHECK-DAG:   %[[C2:.+]] = constant 2 : index
 //   CHECK-DAG:   %[[C3:.+]] = constant 3 : index
 //   CHECK-DAG:   %[[C5:.+]] = constant 5 : index
-//   CHECK-DAG:   %[[D0:.+]] = memref.dim %[[ARG_1]], %[[C0]]
+//   CHECK-DAG:   %[[D0:.+]] = tensor.dim %[[ARG_1]], %[[C0]]
 //   CHECK-DAG:   %[[S0:.+]] = tensor.from_elements %[[D0]], %[[C5]]
 //   CHECK-DAG:   %[[D0_OUT:.+]] = tensor.extract %[[S0]][%[[C0]]]
-//   CHECK-DAG:   %[[D1:.+]] = memref.dim %[[ARG_0]], %[[C2]]
+//   CHECK-DAG:   %[[D1:.+]] = tensor.dim %[[ARG_0]], %[[C2]]
 //   CHECK-DAG:   %[[S1:.+]] = tensor.from_elements %[[C2]], %[[C3]], %[[D1]]
 //   CHECK-DAG:   %[[D1_OUT:.+]] = tensor.extract %[[S1]][%[[C2]]]
 //       CHECK:   return %[[D0_OUT]], %[[C5]], %[[C2]], %[[C3]], %[[D1_OUT]]

mlir/test/Transforms/canonicalize.mlir

func @test_subi_zero_tensor(%arg0: tensor<4x5xi32>) -> tensor<4x5xi32> {
   return %y: tensor<4x5xi32>
 }
 
 // CHECK-LABEL: func @dim
 func @dim(%arg0: tensor<8x4xf32>) -> index {
 
   // CHECK: %c4 = constant 4 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c1 : tensor<8x4xf32>
+  %0 = tensor.dim %arg0, %c1 : tensor<8x4xf32>
 
   // CHECK-NEXT: return %c4
   return %0 : index
 }
 
 // CHECK-LABEL: func @test_commutative
 func @test_commutative(%arg0: i32) -> (i32, i32) {
   // CHECK: %c42_i32 = constant 42 : i32
   %c42_i32 = constant 42 : i32
   // CHECK-NEXT: %0 = addi %arg0, %c42_i32 : i32
   %y = addi %c42_i32, %arg0 : i32
 
   // This should not be swapped.
   // CHECK-NEXT: %1 = subi %c42_i32, %arg0 : i32
   %z = subi %c42_i32, %arg0 : i32
 
   // CHECK-NEXT: return %0, %1
   return %y, %z: i32, i32
 }
 
 // CHECK-LABEL: func @trivial_dce
 func @trivial_dce(%arg0: tensor<8x4xf32>) {
   %c1 = constant 1 : index
-  %0 = memref.dim %arg0, %c1 : tensor<8x4xf32>
+  %0 = tensor.dim %arg0, %c1 : tensor<8x4xf32>
   // CHECK-NEXT: return
   return
 }
 
 // CHECK-LABEL: func @load_dce
 func @load_dce(%arg0: index) {
   %c4 = constant 4 : index
   %a = memref.alloc(%c4) : memref<?xf32>

mlir/test/Transforms/constant-fold.mlir

func @muli_splat_vector() -> vector<4xi32> {
   return %2 : vector<4xi32>
 }
 
 // CHECK-LABEL: func @dim
 func @dim(%x : tensor<8x4xf32>) -> index {
 
   // CHECK:[[C4:%.+]] = constant 4 : index
   %c1 = constant 1 : index
-  %0 = memref.dim %x, %c1 : tensor<8x4xf32>
+  %0 = tensor.dim %x, %c1 : tensor<8x4xf32>
 
   // CHECK-NEXT: return [[C4]]
   return %0 : index
 }
 
 // -----
 
 // CHECK-LABEL: func @cmpi

mlir/test/lib/Dialect/Test/TestDialect.cpp

 //===- TestDialect.cpp - MLIR Dialect for Testing -------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "TestDialect.h"
 #include "TestAttributes.h"
 #include "TestInterfaces.h"
 #include "TestTypes.h"
 #include "mlir/Dialect/DLTI/DLTI.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/DialectImplementation.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/IR/TypeUtilities.h"
 #include "mlir/Reducer/ReductionPatternInterface.h"
 #include "mlir/Transforms/FoldUtils.h"
LogicalResult OpWithShapedTypeInferTypeInterfaceOp::inferReturnTypeComponents(
   inferredReturnShapes.push_back(ShapedTypeComponents({dim}, type));
   return success();
 }
 
 LogicalResult OpWithShapedTypeInferTypeInterfaceOp::reifyReturnTypeShapes(
     OpBuilder &builder, ValueRange operands,
     llvm::SmallVectorImpl<Value> &shapes) {
   shapes = SmallVector<Value, 1>{
-      builder.createOrFold<memref::DimOp>(getLoc(), operands.front(), 0)};
+      builder.createOrFold<tensor::DimOp>(getLoc(), operands.front(), 0)};
   return success();
 }
 
 LogicalResult OpWithResultShapeInterfaceOp::reifyReturnTypeShapes(
     OpBuilder &builder, ValueRange operands,
     llvm::SmallVectorImpl<Value> &shapes) {
   Location loc = getLoc();
   shapes.reserve(operands.size());
   for (Value operand : llvm::reverse(operands)) {
     auto currShape = llvm::to_vector<4>(llvm::map_range(
         llvm::seq<int64_t>(
             0, operand.getType().cast<RankedTensorType>().getRank()),
         [&](int64_t dim) -> Value {
-          return builder.createOrFold<memref::DimOp>(loc, operand, dim);
+          return builder.createOrFold<tensor::DimOp>(loc, operand, dim);
         }));
     shapes.push_back(builder.create<tensor::FromElementsOp>(
         getLoc(), builder.getIndexType(), currShape));
   }
   return success();
 }
 
 LogicalResult
 OpWithResultShapePerDimInterfaceOp ::reifyReturnTypeShapesPerResultDim(
     OpBuilder &builder,
     llvm::SmallVectorImpl<llvm::SmallVector<Value>> &shapes) {
   Location loc = getLoc();
   shapes.reserve(getNumOperands());
   for (Value operand : llvm::reverse(getOperands())) {
     auto currShape = llvm::to_vector<4>(llvm::map_range(
         llvm::seq<int64_t>(
             0, operand.getType().cast<RankedTensorType>().getRank()),
         [&](int64_t dim) -> Value {
-          return builder.createOrFold<memref::DimOp>(loc, operand, dim);
+          return builder.createOrFold<tensor::DimOp>(loc, operand, dim);
         }));
     shapes.emplace_back(std::move(currShape));
   }
   return success();
 }
 
 LogicalResult OpWithResultShapeAndPerDimInterfaceOp::reifyReturnTypeShapes(
     OpBuilder &builder, ValueRange operands,
     llvm::SmallVectorImpl<Value> &shapes) {
   Location loc = getLoc();
   shapes.reserve(operands.size());
   for (Value operand : llvm::reverse(operands)) {
     auto currShape = llvm::to_vector<4>(llvm::map_range(
         llvm::seq<int64_t>(
             0, operand.getType().cast<RankedTensorType>().getRank()),
         [&](int64_t dim) -> Value {
-          return builder.createOrFold<memref::DimOp>(loc, operand, dim);
+          return builder.createOrFold<tensor::DimOp>(loc, operand, dim);
         }));
     shapes.push_back(builder.create<tensor::FromElementsOp>(
         getLoc(), builder.getIndexType(), currShape));
   }
   return success();
 }
 
 LogicalResult
 OpWithResultShapeAndPerDimInterfaceOp ::reifyReturnTypeShapesPerResultDim(
     OpBuilder &builder,
     llvm::SmallVectorImpl<llvm::SmallVector<Value>> &shapes) {
   Location loc = getLoc();
   shapes.reserve(getNumOperands());
   for (Value operand : llvm::reverse(getOperands())) {
     auto currShape = llvm::to_vector<4>(llvm::map_range(
         llvm::seq<int64_t>(
             0, operand.getType().cast<RankedTensorType>().getRank()),
         [&](int64_t dim) -> Value {
-          return builder.createOrFold<memref::DimOp>(loc, operand, dim);
+          return builder.createOrFold<tensor::DimOp>(loc, operand, dim);
         }));
     shapes.emplace_back(std::move(currShape));
   }
   return success();
 }
 
 //===----------------------------------------------------------------------===//
 // Test SideEffect interfaces

mlir/test/lib/Dialect/Test/TestPatterns.cpp

 //===- TestPatterns.cpp - Test dialect pattern driver ---------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "TestDialect.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/StandardOps/Transforms/FuncConversions.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/Matchers.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/FoldUtils.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
 using namespace mlir;
 using namespace mlir::test;
for (auto it : llvm::enumerate(shapes)) {
     op->emitRemark() << "value " << it.index() << ": "
                      << it.value().getDefiningOp();
   }
 }
 
 struct TestReturnTypeDriver
     : public PassWrapper<TestReturnTypeDriver, FunctionPass> {
   void getDependentDialects(DialectRegistry &registry) const override {
-    registry.insert<memref::MemRefDialect>();
+    registry.insert<tensor::TensorDialect>();
   }
   StringRef getArgument() const final { return "test-return-type"; }
   StringRef getDescription() const final { return "Run return type functions"; }
 
   void runOnFunction() override {
     if (getFunction().getName() == "testCreateFunctions") {
       std::vector<Operation *> ops;
       // Collect ops to avoid triggering on inserted ops.