[MLIR] Create memref dialect and move dialect-specific ops from std.

[dfki-jugr]

Summary

Create the memref dialect and move dialect-specific ops
from std dialect to this dialect.

Moved ops:
AllocOp -> MemRef_AllocOp
AllocaOp -> MemRef_AllocaOp
AssumeAlignmentOp -> MemRef_AssumeAlignmentOp
DeallocOp -> MemRef_DeallocOp
DimOp -> MemRef_DimOp
MemRefCastOp -> MemRef_CastOp
MemRefReinterpretCastOp -> MemRef_ReinterpretCastOp
GetGlobalMemRefOp -> MemRef_GetGlobalOp
GlobalMemRefOp -> MemRef_GlobalOp
LoadOp -> MemRef_LoadOp
PrefetchOp -> MemRef_PrefetchOp
ReshapeOp -> MemRef_ReshapeOp
StoreOp -> MemRef_StoreOp
SubViewOp -> MemRef_SubViewOp
TransposeOp -> MemRef_TransposeOp
TensorLoadOp -> MemRef_TensorLoadOp
TensorStoreOp -> MemRef_TensorStoreOp
TensorToMemRefOp -> MemRef_BufferCastOp
ViewOp -> MemRef_ViewOp

The roadmap to split the memref dialect from std is discussed here:
https://llvm.discourse.group/t/rfc-split-the-memref-dialect-from-std/2667

[rriddle]

Do you have a plan for splitting out the memref parts of the StandardTo*/*ToStandard passes?

Thanks for tackling this!

mlir/docs/Dialects/MemRef.md

2–3

This doesn't seem true for MemRef dialect?

mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp

286–287

This comment needs updated, same for the other copies as well. I think this also applies the chapter documentation in docs/, can you check?

mlir/include/mlir/Conversion/Passes.td

319–324

Same with the others?

mlir/include/mlir/Dialect/Linalg/Passes.td

83–87

Is this over 80 characters? Can you reflow if so?

101–106

Please format this.

146–151

Over 80 characters? Please reflow.

mlir/include/mlir/Dialect/MemRef/EDSC/Intrinsics.h

9

Please fix

31

This looks off.

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

13

Do you need all of these?

43

Shouldn't these be in the memref namespace?

80

Unrelated to the current patch, but we should really move these to ODS.

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

19

Nit: Move the description above the hasConstantMaterializer. The description should generally be close to the top of the def.

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

114

This looks off.

133

Is this still necessary? It was only necessary for standard dialect because of the dialect prefix elision.

mlir/include/mlir/Dialect/StandardOps/Utils/Utils.h

34

Can this be forward declared?

39

Drop the llvm::

mlir/include/mlir/Transforms/Utils.h

95

Can you forward declare this op instead of including the above file?

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

191

Drop trivial braces.

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

17 ↗

(On Diff #328501)

Why is this one needed?

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

24

Can you add this dependency in the ODS definition instead?

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

20

Add this in ODS instead?

mlir/lib/Transforms/Canonicalizer.cpp

35

The canonicalizer should not have a dependency on any dialect, please remove this.

[herhut]

Wow, what a patch :)

Some nits but otherwise looks good.

mlir/docs/Rationale/UsageOfConst.md

204

I think this change should be undone. Or you need to replace standard with memref, as well.

mlir/docs/Traits.md

214

The link will be broken.

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

116

This has only one user it seems. Maybe just drop it and inline to the single use?

418

So now all dim operations live in the memref dialect, independently of whether they are on memref or tensor? I assume this is temporary until we find a better solution?

699

hyper-nit: spaces around = while you are here.

mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp

21

Why is this needed?

mlir/lib/Dialect/Affine/Transforms/AffineDataCopyGeneration.cpp

26

Why is this needed?

mlir/lib/Dialect/GPU/Transforms/MemoryPromotion.cpp

16

Extra include without any other changes?

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

25–26

Please move to include/mlir/Dialect/StandardOps/Transforms/Passes.td

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

104

Please move to include/mlir/Dialect/StandardOps/Transforms/Passes.td

mlir/lib/Transforms/Bufferize.cpp

73

nit: fix name.

[dfki-jugr]

Adressed comments.

[herhut]

Please address the clang tidy warnings. Otherwise this looks good to go to me.

[rriddle]

mlir/lib/Transforms/Canonicalizer.cpp

15

Remove this.

[bondhugula]

mlir/include/mlir/Transforms/Utils.h

95

@dfki-jugr Why does this revision change the signature of this function here to pass AllocOp by pointer? As far as I can tell from the summary and a quick, this is only about the movement of ops as opposed to an API change.

[ftynse]

mlir/include/mlir/Transforms/Utils.h

95

Because it can then use a forward declaration instead of having MLIRTransforms depend on MLIRMemRef?

[mehdi_amini]

mlir/lib/Transforms/Utils/Utils.cpp

20

@ftynse: Won't libTransforms depend on memref anyway here?

[silvas]

It looks like this patch simply moved std.dim to memref.dim, without doing the "splitting" which we had discussed. I now have code that only operates on tensors that is needing to pull in the memref dialect just for this. Can you split memref.dim into tensor.dim for the tensor case?

[dfki-jugr]

In D98041#2645874, @silvas wrote:

It looks like this patch simply moved std.dim to memref.dim, without doing the "splitting" which we had discussed. I now have code that only operates on tensors that is needing to pull in the memref dialect just for this. Can you split memref.dim into tensor.dim for the tensor case?

This is true. Since it was up for discussion, how to handle such a splitting of ops (https://llvm.discourse.group/t/cross-dialect-folding-and-canonicalization/2740/11), the dim op is moved completely to the memref dialect. We are looking forward to an agreement on how to split the op.

[silvas]

I think just a patch splitting the op into tensor.dim and memref.dim would be fine. There isn't much to discuss.

[herhut]

In D98041#2648155, @silvas wrote:

I think just a patch splitting the op into tensor.dim and memref.dim would be fine. There isn't much to discuss.

So there was a discussion that this could be done via interfaces, for example. We could have a generic dim and rank operation that can be applied to any ShapedType type. That would also make it easier to rewrite between these types, as the dim and rank operation would no longer need to change.

Just to be clear, I am not saying we should not do this split, I am just trying to avoid the extra work if soon later we decide that there is a different and better way to model this. This is more work than "just a patch", especially for downstream users, so we should avoid unnecessary churn. As far as I understand it, this is about an extra link time dependency that is not worse than the monolithic dialect before.

[silvas]

In D98041#2650101, @herhut wrote:

In D98041#2648155, @silvas wrote:

I think just a patch splitting the op into tensor.dim and memref.dim would be fine. There isn't much to discuss.

So there was a discussion that this could be done via interfaces, for example. We could have a generic dim and rank operation that can be applied to any ShapedType type. That would also make it easier to rewrite between these types, as the dim and rank operation would no longer need to change.

Just to be clear, I am not saying we should not do this split, I am just trying to avoid the extra work if soon later we decide that there is a different and better way to model this. This is more work than "just a patch", especially for downstream users, so we should avoid unnecessary churn. As far as I understand it, this is about an extra link time dependency that is not worse than the monolithic dialect before.

ShapedType only covers tensor and memref. I'm not currently aware of any effort to generalize that, so splitting it now seems fairly safe.

I'm also well aware of the cost to downstream users, as I just updated my own codebases to memref.dim. The reason I wanted it to be split in the first place is to reduce that. Now when we split memref.dim and tensor.dim all existing code will compile and just have verifier errors... if we had initially done the split std.dim -> {tensor.dim,memref.dim} then we would have only had compile errors. I'm sorry I didn't notice this on the original patch or I would have mentioned it.

[silvas]

Any progress on splitting memref.dim into tensor.dim? Multiple downstream projects are now complaining about this.

[bondhugula]

In D98041#2651574, @silvas wrote:

In D98041#2650101, @herhut wrote:

In D98041#2648155, @silvas wrote:

I think just a patch splitting the op into tensor.dim and memref.dim would be fine. There isn't much to discuss.

So there was a discussion that this could be done via interfaces, for example. We could have a generic dim and rank operation that can be applied to any ShapedType type. That would also make it easier to rewrite between these types, as the dim and rank operation would no longer need to change.

Just to be clear, I am not saying we should not do this split, I am just trying to avoid the extra work if soon later we decide that there is a different and better way to model this. This is more work than "just a patch", especially for downstream users, so we should avoid unnecessary churn. As far as I understand it, this is about an extra link time dependency that is not worse than the monolithic dialect before.

ShapedType only covers tensor and memref. I'm not currently aware of any effort to generalize that, so splitting it now seems fairly safe.

Not really. VectorType is also a ShapedType, but the current dim op works only on tensors and memrefs (since vectors being of static shape would always have a dim on them folding to a constant).

It is odd now that memref.dim is needed for tensors as well. The current op documentation is also inaccurate now - the definition says "any memref or tensor type" but the description says memref only.

def DimOp : MemRef_Op<"dim", [NoSideEffect]> {
  let summary = "dimension index operation";
  let description = [{
    The `dim` operation takes a memref and a dimension operand of type `index`.
  ...
 let arguments = (ins AnyTypeOf<[AnyTensor, AnyRankedOrUnrankedMemRef],
                                 "any memref or tensor type">:$memrefOrTensor,
                       Index:$index);

mlir/include/mlir/Transforms/Utils.h

95

MLIRTransforms already depends on memrefs - but I guess you meant avoiding a header include. Passing a derived op type by pointer creates confusion to the extent that it might be a nullable argument which is definitely not the case here.

[mehdi_amini]

mlir/include/mlir/Transforms/Passes.td

366

That is looking really suspicious: the canonicalize pass itself should not have any such dependency. It also seems related to River's comment.

mlir/include/mlir/Transforms/Utils.h

95

@dfki-jugr : this comment wasn't addressed either.

mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp

21

This comment wasn't addressed.

mlir/lib/Dialect/Affine/Transforms/AffineDataCopyGeneration.cpp

26

ditto.

mlir/lib/Transforms/Canonicalizer.cpp

15

This wasn't done apparently?

35

This was moved to TableGen instead of being removed ....

[mehdi_amini]

@dfki-jugr can you acknowledge? (and address...)

[dfki-jugr]

@mehdi_amini We have already discussed this "issue" with @herhut and we detected that the includes can not be removed in a straight forward way. It looks like removing those includes causes memref dependencies that can not be resolved properly. Unfortunately, we are not experts in this particular field of MLIR core. However, we are definitely looking forward to address your concerns. If we are able to get some support regarding these implicitely existing memref dependencies.

mlir/include/mlir/Transforms/Utils.h

95

As @rriddle mentioned, we added a forward declaration instead of an include. We agree that the signature change is misleading. This can be resolved by reverting this change. This should be possible since @bondhugula mentioned that there is already a dependency.

[silvas]

Any progress on splitting memref.dim into tensor.dim?

[mehdi_amini]

@dfki-jugr: Can you specifically comment on why the canonicalize pass depends on the memref dialect? Maybe we can start here: both River and I asked inline about *why* you added this dependency to the canonicalize pass.

[mehdi_amini]

mlir/include/mlir/Transforms/Passes.td

366

@dfki-jugr here in particular if you missed it before

mlir/lib/Transforms/Canonicalizer.cpp

15

@dfki-jugr here in particular if you missed it before, please remove this header, if you can't and need help, please explain exactly why you need this here.

[dfki-jugr]

@mehdi_amini We tried to resolve the dependency, but ran into the problem that the canonicalize test did not recognize the memref ops. Before splitting the dialect, this was not an issue since the ops were part of the std-dialect. Now, we get an error that the splitted memref ops are not registered in MLIR context. That's the reason why this dependency is needed. As mentioned above, we are not experts in this particular field of MLIR and are looking forward to get some support how to get rid of this dependency. Maybe we can pre-register the memref-dialect somehow?

[dfki-jugr]

In D98041#2740418, @silvas wrote:

Any progress on splitting memref.dim into tensor.dim?

We are still waiting for an agreement in the discussion how to proceed:

https://llvm.discourse.group/t/cross-dialect-folding-and-canonicalization/2740/13?u=dfki-jugr

[mehdi_amini]

In D98041#2741534, @dfki-jugr wrote:

@mehdi_amini We tried to resolve the dependency, but ran into the problem that the canonicalize test did not recognize the memref ops. Before splitting the dialect, this was not an issue since the ops were part of the std-dialect. Now, we get an error that the splitted memref ops are not registered in MLIR context. That's the reason why this dependency is needed. As mentioned above, we are not experts in this particular field of MLIR and are looking forward to get some support how to get rid of this dependency. Maybe we can pre-register the memref-dialect somehow?

canonicalize is a generic pass that is used in many situation where memref is not used, for example TensorFlow. I can't see a reason for the Memref dialect to be loaded in the context when I run the canonicalize on TensorFlow, however it is the case after your change here.

You are mentioning that "tried to resolve the dependency, but ran into the problem that the canonicalize test did not recognize the memref ops", can you post the error? (which test and what failed?)
The canonicalize pass itself is just applying patterns provided by dialects, if something is failing it is because a dialect does not implement a dependency on the memref dialect (by loading it in its constructor).

[dfki-jugr]

@mehdi_amini Try to fix the issue here: https://reviews.llvm.org/D102060

[silvas]

Hi @dfki-jugr, can we please split memref.dim into tensor.dim (and std.rank into tensor.rank if you have time). I've now independently heard complaints about this from 5+ people across many different teams I interact with.

mlir/docs/Dialects/MemRef.md

-# 'std' Dialect
+# 'memref' Dialect
 
-This dialect provides documentation for operations within the Standard dialect.
-
-Note: This dialect is a collection of operations for several different concepts,
-and should be split into multiple more-focused dialects accordingly.
+This dialect provides documentation for operations within the MemRef dialect.
 
 **Please post an RFC on the [forum](https://llvm.discourse.group/c/mlir/31)
 before adding or changing any operation in this dialect.**
 
 [TOC]
 
 ## Operations
 
-[include "Dialects/StandardOps.md"]
+[include "Dialects/MemRefOps.md"]
 
 ### 'dma_start' operation
 
 Syntax:
 
 ```
 operation ::= `dma_start` ssa-use`[`ssa-use-list`]` `,`
                ssa-use`[`ssa-use-list`]` `,` ssa-use `,`

mlir/docs/Rationale/UsageOfConst.md

 operation isn't modified) this is a violation of the model, since the returned
 operation must be mutable, and provides access to the full graph of operands as
 the original operation, violating the graph based const model we were shooting
 for.
 
 ### The `OpPointer` and `ConstOpPointer` Classes
 
 The "typed operation" classes for registered operations (e.g. like `DimOp` for
-the "std.dim" operation in standard ops) contain a pointer to an operation and
+the "memref.dim" operation in memref ops) contain a pointer to an operation and
 provide typed APIs for processing it.
 
 However, this is a problem for our current `const` design - `const DimOp` means
 the pointer itself is immutable, not the pointee. The current solution for this
 is the `OpPointer<>` and `ConstOpPointer<>` classes, which exist solely to
 provide const correctness when referring to a typed operation. Instead of
 referring to `DimOp` directly, we need to use `OpPointer<DimOp>` and
 `ConstOpPointer<DimOp>` to preserve this constness.
 

mlir/docs/Traits.md

 
 ### AutomaticAllocationScope
 
 *   `OpTrait::AutomaticAllocationScope` -- `AutomaticAllocationScope`
 
 This trait is carried by region holding operations that define a new scope for
 automatic allocation. Such allocations are automatically freed when control is
 transferred back from the regions of such operations. As an example, allocations
-performed by [`std.alloca`](Dialects/Standard.md#stdalloca-allocaop) are
+performed by [`memref.alloca`](Dialects/MemRef.md#memrefalloca-allocaop) are
 automatically freed when control leaves the region of its closest surrounding op
 that has the trait AutomaticAllocationScope.
 
 ### Broadcastable
 
 *   `OpTrait::ResultsBroadcastableShape` -- `ResultsBroadcastableShape`
 
 This trait adds the property that the operation is known to have

mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp

 //====- LowerToAffineLoops.cpp - Partial lowering from Toy to Affine+Std --===//
 //
 // 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 file implements a partial lowering of Toy operations to a combination of
-// affine loops and standard operations. This lowering expects that all calls
-// have been inlined, and all shapes have been resolved.
+// affine loops, memref operations and standard operations. This lowering
+// expects that all calls have been inlined, and all shapes have been resolved.
 //
 //===----------------------------------------------------------------------===//
 
 #include "toy/Dialect.h"
 #include "toy/Passes.h"
 
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "llvm/ADT/Sequence.h"
 
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
 // ToyToAffine RewritePatterns
 //===----------------------------------------------------------------------===//
 
 /// Convert the given TensorType into the corresponding MemRefType.
 static MemRefType convertTensorToMemRef(TensorType type) {
   assert(type.hasRank() && "expected only ranked shapes");
   return MemRefType::get(type.getShape(), type.getElementType());
 }
 
 /// Insert an allocation and deallocation for the given MemRefType.
 static Value insertAllocAndDealloc(MemRefType type, Location loc,
                                    PatternRewriter &rewriter) {
-  auto alloc = rewriter.create<AllocOp>(loc, type);
+  auto alloc = rewriter.create<memref::AllocOp>(loc, type);
 
   // Make sure to allocate at the beginning of the block.
   auto *parentBlock = alloc->getBlock();
   alloc->moveBefore(&parentBlock->front());
 
   // Make sure to deallocate this alloc at the end of the block. This is fine
   // as toy functions have no control flow.
-  auto dealloc = rewriter.create<DeallocOp>(loc, alloc);
+  auto dealloc = rewriter.create<memref::DeallocOp>(loc, alloc);
   dealloc->moveBefore(&parentBlock->back());
   return alloc;
 }
 
 /// This defines the function type used to process an iteration of a lowered
 /// loop. It takes as input an OpBuilder, an range of memRefOperands
 /// corresponding to the operands of the input operation, and the range of loop
 /// induction variables for the iteration. It returns a value to store at the
LogicalResult matchAndRewrite(toy::ConstantOp op,
     // We will be generating constant indices up-to the largest dimension.
     // Create these constants up-front to avoid large amounts of redundant
     // operations.
     auto valueShape = memRefType.getShape();
     SmallVector<Value, 8> constantIndices;
 
     if (!valueShape.empty()) {
       for (auto i : llvm::seq<int64_t>(
                0, *std::max_element(valueShape.begin(), valueShape.end())))
         constantIndices.push_back(rewriter.create<ConstantIndexOp>(loc, i));
     } else {
       // This is the case of a tensor of rank 0.
       constantIndices.push_back(rewriter.create<ConstantIndexOp>(loc, 0));
     }
 
     // The constant operation represents a multi-dimensional constant, so we
     // will need to generate a store for each of the elements. The following
     // functor recursively walks the dimensions of the constant shape,

 
 /// This is a partial lowering to affine loops of the toy operations that are
 /// computationally intensive (like matmul for example...) while keeping the
 /// rest of the code in the Toy dialect.
 namespace {
 struct ToyToAffineLoweringPass
     : public PassWrapper<ToyToAffineLoweringPass, FunctionPass> {
   void getDependentDialects(DialectRegistry &registry) const override {
-    registry.insert<AffineDialect, StandardOpsDialect>();
+    registry.insert<AffineDialect, memref::MemRefDialect, StandardOpsDialect>();
   }
   void runOnFunction() final;
 };
 } // end anonymous namespace.
 
 void ToyToAffineLoweringPass::runOnFunction() {
   auto function = getFunction();
 
   // We only lower the main function as we expect that all other functions have
   // been inlined.
   if (function.getName() != "main")
     return;
 
   // Verify that the given main has no inputs and results.
   if (function.getNumArguments() || function.getType().getNumResults()) {
     function.emitError("expected 'main' to have 0 inputs and 0 results");
     return signalPassFailure();
   }
 
   // The first thing to define is the conversion target. This will define the
   // final target for this lowering.
   ConversionTarget target(getContext());
 
   // We define the specific operations, or dialects, that are legal targets for
   // this lowering. In our case, we are lowering to a combination of the
-  // `Affine` and `Standard` dialects.
-  target.addLegalDialect<AffineDialect, StandardOpsDialect>();
+  // `Affine`, `MemRef` and `Standard` dialects.
+  target.addLegalDialect<AffineDialect, memref::MemRefDialect,
+                         StandardOpsDialect>();
 
   // We also define the Toy dialect as Illegal so that the conversion will fail
   // if any of these operations are *not* converted. Given that we actually want
   // a partial lowering, we explicitly mark the Toy operations that don't want
   // to lower, `toy.print`, as `legal`.
   target.addIllegalDialect<toy::ToyDialect>();
   target.addLegalOp<toy::PrintOp>();
 

mlir/include/mlir/Conversion/Passes.td

 
 //===----------------------------------------------------------------------===//
 // GPUToNVVM
 //===----------------------------------------------------------------------===//
 
 def ConvertGpuOpsToNVVMOps : Pass<"convert-gpu-to-nvvm", "gpu::GPUModuleOp"> {
   let summary = "Generate NVVM operations for gpu operations";
   let constructor = "mlir::createLowerGpuOpsToNVVMOpsPass()";
-  let dependentDialects = ["NVVM::NVVMDialect"];
+  let dependentDialects = ["NVVM::NVVMDialect", "memref::MemRefDialect"];
   let options = [
     Option<"indexBitwidth", "index-bitwidth", "unsigned",
            /*default=kDeriveIndexBitwidthFromDataLayout*/"0",
            "Bitwidth of the index type, 0 to use size of machine word">
   ];
 }
 
 //===----------------------------------------------------------------------===//

 //===----------------------------------------------------------------------===//
 // LinalgToStandard
 //===----------------------------------------------------------------------===//
 
 def ConvertLinalgToStandard : Pass<"convert-linalg-to-std", "ModuleOp"> {
   let summary = "Convert the operations from the linalg dialect into the "
                 "Standard dialect";
   let constructor = "mlir::createConvertLinalgToStandardPass()";
-  let dependentDialects = ["StandardOpsDialect"];
+  let dependentDialects = ["memref::MemRefDialect", "StandardOpsDialect"];
 }
 
 //===----------------------------------------------------------------------===//
 // LinalgToSPIRV
 //===----------------------------------------------------------------------===//
 
 def ConvertLinalgToSPIRV : Pass<"convert-linalg-to-spirv", "ModuleOp"> {
   let summary = "Convert Linalg dialect to SPIR-V dialect";

 //===----------------------------------------------------------------------===//
 // 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 = ["StandardOpsDialect", "scf::SCFDialect"];
+  let dependentDialects = [
+    "memref::MemRefDialect",
+    "StandardOpsDialect",
+    "scf::SCFDialect"
+  ];
 }
 
 def ConvertShapeConstraints: Pass<"convert-shape-constraints", "FuncOp"> {
   let summary = "Convert shape constraint operations to the standard dialect";
   let description = [{
     This pass eliminates shape constraints from the program, converting them to
     eager (side-effecting) error handling code.
 
     This pass is separate from the regular convert-shape-to-standard, despite

 //===----------------------------------------------------------------------===//
 // VectorToSCF
 //===----------------------------------------------------------------------===//
 
 def ConvertVectorToSCF : FunctionPass<"convert-vector-to-scf"> {
   let summary = "Lower the operations from the vector dialect into the SCF "
                 "dialect";
   let constructor = "mlir::createConvertVectorToSCFPass()";
-  let dependentDialects = ["AffineDialect", "scf::SCFDialect"];
+  let dependentDialects = [
+    "AffineDialect",
+    "memref::MemRefDialect",
+    "scf::SCFDialect"
+  ];
   let options = [
     Option<"fullUnroll", "full-unroll", "bool", /*default=*/"false",
            "Perform full unrolling when converting vector transfers to SCF">,
   ];
 }
 
 //===----------------------------------------------------------------------===//
 // VectorToLLVM

mlir/include/mlir/Dialect/Linalg/Passes.td

 def ConvertElementwiseToLinalg : FunctionPass<"convert-elementwise-to-linalg"> {
   let summary = "Convert ElementwiseMappable ops to linalg";
   let description = [{
     Convert ops with the `ElementwiseMappable` trait to linalg parallel loops.
 
     This pass only converts ops that operate on ranked tensors.
   }];
   let constructor = "mlir::createConvertElementwiseToLinalgPass()";
-  let dependentDialects = ["linalg::LinalgDialect"];
+  let dependentDialects = ["linalg::LinalgDialect", "memref::MemRefDialect"];
 }
 
 def LinalgFoldUnitExtentDims : FunctionPass<"linalg-fold-unit-extent-dims"> {
   let summary = "Remove unit-extent dimension in Linalg ops on tensors";
   let constructor = "mlir::createLinalgFoldUnitExtentDimsPass()";
   let options = [
     Option<"foldOneTripLoopsOnly", "fold-one-trip-loops-only", "bool",
             /*default=*/"false",
def LinalgLowerToLoops : FunctionPass<"convert-linalg-to-loops"> {
   let summary = "Lower the operations from the linalg dialect into loops";
   let constructor = "mlir::createConvertLinalgToLoopsPass()";
   let options = [
     ListOption<"interchangeVector", "interchange-vector", "unsigned",
                "Permute the loops in the nest following the given "
                "interchange vector",
                "llvm::cl::ZeroOrMore, llvm::cl::MiscFlags::CommaSeparated">
   ];
-  let dependentDialects = ["linalg::LinalgDialect", "scf::SCFDialect", "AffineDialect"];
+  let dependentDialects = [
+    "linalg::LinalgDialect",
+    "scf::SCFDialect",
+    "AffineDialect"
+  ];
 }
 
 def LinalgBufferize : Pass<"linalg-bufferize", "FuncOp"> {
   let summary = "Bufferize the linalg dialect";
   let constructor = "mlir::createLinalgBufferizePass()";
-  let dependentDialects = ["linalg::LinalgDialect", "AffineDialect"];
+  let dependentDialects = [
+    "linalg::LinalgDialect",
+    "AffineDialect",
+    "memref::MemRefDialect"
+  ];
 }
 
 def LinalgLowerToParallelLoops
     : FunctionPass<"convert-linalg-to-parallel-loops"> {
   let summary = "Lower the operations from the linalg dialect into parallel "
                 "loops";
   let constructor = "mlir::createConvertLinalgToParallelLoopsPass()";
   let options = [
     ListOption<"interchangeVector", "interchange-vector", "unsigned",
                "Permute the loops in the nest following the given "
                "interchange vector",
                "llvm::cl::ZeroOrMore, llvm::cl::MiscFlags::CommaSeparated">
   ];
-  let dependentDialects = ["AffineDialect", "linalg::LinalgDialect", "scf::SCFDialect"];
+  let dependentDialects = [
+    "AffineDialect",
+    "linalg::LinalgDialect",
+    "memref::MemRefDialect",
+    "scf::SCFDialect"
+  ];
 }
 
 def LinalgPromotion : FunctionPass<"linalg-promote-subviews"> {
   let summary = "Promote subview ops to local buffers";
   let constructor = "mlir::createLinalgPromotionPass()";
   let options = [
     Option<"dynamicBuffers", "test-promote-dynamic", "bool",
            /*default=*/"false", "Test generation of dynamic promoted buffers">,
     Option<"useAlloca", "test-use-alloca", "bool",
            /*default=*/"false", "Test generation of alloca'ed buffers.">
   ];
   let dependentDialects = ["linalg::LinalgDialect"];
 }
 
 def LinalgTiling : FunctionPass<"linalg-tile"> {
   let summary = "Tile operations in the linalg dialect";
   let constructor = "mlir::createLinalgTilingPass()";
   let dependentDialects = [
-    "AffineDialect", "linalg::LinalgDialect", "scf::SCFDialect"
+    "AffineDialect",
+    "linalg::LinalgDialect",
+    "memref::MemRefDialect",
+    "scf::SCFDialect"
   ];
   let options = [
     ListOption<"tileSizes", "linalg-tile-sizes", "int64_t",
                "Test generation of dynamic promoted buffers",
                "llvm::cl::ZeroOrMore, llvm::cl::MiscFlags::CommaSeparated">
   ];
 }
 
 def LinalgTilingToParallelLoops
     : FunctionPass<"linalg-tile-to-parallel-loops"> {
   let summary = "Tile operations in the linalg dialect to parallel loops";
   let constructor = "mlir::createLinalgTilingToParallelLoopsPass()";
   let options = [
     ListOption<"tileSizes", "linalg-tile-sizes", "int64_t",
                "Test generation of dynamic promoted buffers",
                "llvm::cl::ZeroOrMore, llvm::cl::MiscFlags::CommaSeparated">
   ];
-  let dependentDialects = ["AffineDialect", "linalg::LinalgDialect", "scf::SCFDialect"];
+  let dependentDialects = [
+    "AffineDialect",
+    "linalg::LinalgDialect",
+    "memref::MemRefDialect",
+    "scf::SCFDialect"
+  ];
 }
 
 def LinalgGeneralization : FunctionPass<"linalg-generalize-named-ops"> {
   let summary = "Convert named ops into generic ops";
   let constructor = "mlir::createLinalgGeneralizationPass()";
   let dependentDialects = ["linalg::LinalgDialect"];
 }
 

mlir/include/mlir/Dialect/MemRef/EDSC/Intrinsics.h

+//===- Intrinsics.h - MLIR EDSC Intrinsics for MemRefOps --------*- 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 MLIR_DIALECT_MEMREF_EDSC_INTRINSICS_H_
+#define MLIR_DIALECT_MEMREF_EDSC_INTRINSICS_H_
+
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/EDSC/Builders.h"
+
+namespace mlir {
+namespace edsc {
+namespace intrinsics {
+
+using memref_alloc = ValueBuilder<memref::AllocOp>;
+using memref_alloca = ValueBuilder<memref::AllocaOp>;
+using memref_cast = ValueBuilder<memref::CastOp>;
+using memref_dealloc = OperationBuilder<memref::DeallocOp>;
+using memref_dim = ValueBuilder<memref::DimOp>;
+using memref_load = ValueBuilder<memref::LoadOp>;
+using memref_store = OperationBuilder<memref::StoreOp>;
+using memref_sub_view = ValueBuilder<memref::SubViewOp>;
+using memref_tensor_load = ValueBuilder<memref::TensorLoadOp>;
+using memref_tensor_store = OperationBuilder<memref::TensorStoreOp>;
+using memref_view = ValueBuilder<memref::ViewOp>;
+
+/// Provide an index notation around memref_load and memref_store.
+using MemRefIndexedValue =
+    TemplatedIndexedValue<intrinsics::memref_load, intrinsics::memref_store>;
+} // namespace intrinsics
+} // namespace edsc
+} // namespace mlir
+
+#endif // MLIR_DIALECT_MEMREF_EDSC_INTRINSICS_H_

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

-//===- Ops.h - Standard MLIR Operations -------------------------*- C++ -*-===//
+//===- MemRef.h - MemRef dialect --------------------------------*- 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
 //
 //===----------------------------------------------------------------------===//
-//
-// This file defines convenience types for working with standard operations
-// in the MLIR operation set.
-//
-//===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_STANDARDOPS_IR_OPS_H
-#define MLIR_DIALECT_STANDARDOPS_IR_OPS_H
+#ifndef MLIR_DIALECT_MEMREF_IR_MEMREF_H_
+#define MLIR_DIALECT_MEMREF_IR_MEMREF_H_
 
-#include "mlir/IR/Builders.h"
-#include "mlir/IR/BuiltinTypes.h"
 #include "mlir/IR/Dialect.h"
-#include "mlir/IR/OpImplementation.h"
 #include "mlir/Interfaces/CallInterfaces.h"
 #include "mlir/Interfaces/CastInterfaces.h"
-#include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
-#include "mlir/Interfaces/VectorInterfaces.h"
 #include "mlir/Interfaces/ViewLikeInterface.h"
 
-// Pull in all enum type definitions and utility function declarations.
-#include "mlir/Dialect/StandardOps/IR/OpsEnums.h.inc"
-
 namespace mlir {
-class AffineMap;
-class Builder;
-class FuncOp;
-class OpBuilder;
-
 raw_ostream &operator<<(raw_ostream &os, Range &range);
 
 /// Return the list of Range (i.e. offset, size, stride). Each Range
 /// 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);
+} // namespace mlir
 
-#define GET_OP_CLASSES
-#include "mlir/Dialect/StandardOps/IR/Ops.h.inc"
-
-#include "mlir/Dialect/StandardOps/IR/OpsDialect.h.inc"
-
-/// This is a refinement of the "constant" op for the case where it is
-/// returning a float value of FloatType.
-///
-///   %1 = "std.constant"(){value: 42.0} : bf16
-///
-class ConstantFloatOp : public ConstantOp {
-public:
-  using ConstantOp::ConstantOp;
-
-  /// Builds a constant float op producing a float of the specified type.
-  static void build(OpBuilder &builder, OperationState &result,
-                    const APFloat &value, FloatType type);
-
-  APFloat getValue() {
-    return (*this)->getAttrOfType<FloatAttr>("value").getValue();
-  }
-
-  static bool classof(Operation *op);
-};
-
-/// This is a refinement of the "constant" op for the case where it is
-/// returning an integer value of IntegerType.
-///
-///   %1 = "std.constant"(){value: 42} : i32
-///
-class ConstantIntOp : public ConstantOp {
-public:
-  using ConstantOp::ConstantOp;
-  /// Build a constant int op producing an integer of the specified width.
-  static void build(OpBuilder &builder, OperationState &result, int64_t value,
-                    unsigned width);
-
-  /// Build a constant int op producing an integer with the specified type,
-  /// which must be an integer type.
-  static void build(OpBuilder &builder, OperationState &result, int64_t value,
-                    Type type);
-
-  int64_t getValue() {
-    return (*this)->getAttrOfType<IntegerAttr>("value").getInt();
-  }
-
-  static bool classof(Operation *op);
-};
-
-/// This is a refinement of the "constant" op for the case where it is
-/// returning an integer value of Index type.
-///
-///   %1 = "std.constant"(){value: 99} : () -> index
-///
-class ConstantIndexOp : public ConstantOp {
-public:
-  using ConstantOp::ConstantOp;
+//===----------------------------------------------------------------------===//
+// MemRef Dialect
+//===----------------------------------------------------------------------===//
 
-  /// Build a constant int op producing an index.
-  static void build(OpBuilder &builder, OperationState &result, int64_t value);
+#include "mlir/Dialect/MemRef/IR/MemRefOpsDialect.h.inc"
 
-  int64_t getValue() {
-    return (*this)->getAttrOfType<IntegerAttr>("value").getInt();
-  }
+//===----------------------------------------------------------------------===//
+// MemRef Dialect Operations
+//===----------------------------------------------------------------------===//
 
-  static bool classof(Operation *op);
-};
+#define GET_OP_CLASSES
+#include "mlir/Dialect/MemRef/IR/MemRefOps.h.inc"
 
+namespace mlir {
+namespace memref {
 // DmaStartOp starts a non-blocking DMA operation that transfers data from a
 // source memref to a destination memref. The source and destination memref need
 // not be of the same dimensionality, but need to have the same elemental type.
 // The operands include the source and destination memref's each followed by its
 // indices, size of the data transfer in terms of the number of elements (of the
 // elemental type of the memref), a tag memref with its indices, and optionally
 // at the end, a stride and a number_of_elements_per_stride arguments. The tag
 // location is used by a DmaWaitOp to check for completion. The indices of the
 // source memref, destination memref, and the tag memref have the same

 //   memory space 0 until %num_elements are transferred.
 //
 //   dma_start %src[%i, %j], %dst[%k, %l], %num_elements, %tag[%idx], %stride,
 //             %num_elt_per_stride :
 //
 // TODO: add additional operands to allow source and destination striding, and
 // multiple stride levels.
 // TODO: Consider replacing src/dst memref indices with view memrefs.
 class DmaStartOp
     : public Op<DmaStartOp, OpTrait::VariadicOperands, OpTrait::ZeroResult> {
 public:
   using Op::Op;
 
   static void build(OpBuilder &builder, OperationState &result, Value srcMemRef,
                     ValueRange srcIndices, Value destMemRef,
                     ValueRange destIndices, Value numElements, Value tagMemRef,
                     ValueRange tagIndices, Value stride = nullptr,
public:
   /// Given a DMA start operation, returns the operand position of either the
   /// source or destination memref depending on the one that is at the higher
   /// level of the memory hierarchy. Asserts failure if neither is true.
   unsigned getFasterMemPos() {
     assert(isSrcMemorySpaceFaster() || isDestMemorySpaceFaster());
     return isSrcMemorySpaceFaster() ? 0 : getSrcMemRefRank() + 1;
   }
 
-  static StringRef getOperationName() { return "std.dma_start"; }
+  static StringRef getOperationName() { return "memref.dma_start"; }
   static ParseResult parse(OpAsmParser &parser, OperationState &result);
   void print(OpAsmPrinter &p);
   LogicalResult verify();
 
   LogicalResult fold(ArrayRef<Attribute> cstOperands,
                      SmallVectorImpl<OpFoldResult> &results);
 
   bool isStrided() {

 class DmaWaitOp
     : public Op<DmaWaitOp, OpTrait::VariadicOperands, OpTrait::ZeroResult> {
 public:
   using Op::Op;
 
   static void build(OpBuilder &builder, OperationState &result, Value tagMemRef,
                     ValueRange tagIndices, Value numElements);
 
-  static StringRef getOperationName() { return "std.dma_wait"; }
+  static StringRef getOperationName() { return "memref.dma_wait"; }
 
   // Returns the Tag MemRef associated with the DMA operation being waited on.
   Value getTagMemRef() { return getOperand(0); }
 
   // Returns the tag memref index for this DMA operation.
   operand_range getTagIndices() {
     return {(*this)->operand_begin() + 1,
             (*this)->operand_begin() + 1 + getTagMemRefRank()};
   }
 
   // Returns the rank (number of indices) of the tag memref.
   unsigned getTagMemRefRank() {
     return getTagMemRef().getType().cast<MemRefType>().getRank();
   }
 
   // Returns the number of elements transferred in the associated DMA operation.
   Value getNumElements() { return getOperand(1 + getTagMemRefRank()); }
 
   static ParseResult parse(OpAsmParser &parser, OperationState &result);
   void print(OpAsmPrinter &p);
   LogicalResult fold(ArrayRef<Attribute> cstOperands,
                      SmallVectorImpl<OpFoldResult> &results);
   LogicalResult verify();
 };
+} // namespace memref
+} // namespace mlir
 
-/// Given an `originalShape` and a `reducedShape` assumed to be a subset of
-/// `originalShape` with some `1` entries erased, return the set of indices
-/// that specifies which of the entries of `originalShape` are dropped to obtain
-/// `reducedShape`. The returned mask can be applied as a projection to
-/// `originalShape` to obtain the `reducedShape`. This mask is useful to track
-/// which dimensions must be kept when e.g. compute MemRef strides under
-/// rank-reducing operations. Return None if reducedShape cannot be obtained
-/// by dropping only `1` entries in `originalShape`.
-llvm::Optional<llvm::SmallDenseSet<unsigned>>
-computeRankReductionMask(ArrayRef<int64_t> originalShape,
-                         ArrayRef<int64_t> reducedShape);
-
-/// Determines whether MemRefCastOp casts to a more dynamic version of the
-/// source memref. This is useful to to fold a memref_cast into a consuming op
-/// and implement canonicalization patterns for ops in different dialects that
-/// may consume the results of memref_cast operations. Such foldable memref_cast
-/// operations are typically inserted as `view` and `subview` ops and are
-/// canonicalized, to preserve the type compatibility of their uses.
-///
-/// Returns true when all conditions are met:
-/// 1. source and result are ranked memrefs with strided semantics and same
-/// element type and rank.
-/// 2. each of the source's size, offset or stride has more static information
-/// than the corresponding result's size, offset or stride.
-///
-/// Example 1:
-/// ```mlir
-///   %1 = memref_cast %0 : memref<8x16xf32> to memref<?x?xf32>
-///   %2 = consumer %1 ... : memref<?x?xf32> ...
-/// ```
-///
-/// may fold into:
-///
-/// ```mlir
-///   %2 = consumer %0 ... : memref<8x16xf32> ...
-/// ```
-///
-/// Example 2:
-/// ```
-///   %1 = memref_cast %0 : memref<?x16xf32, affine_map<(i, j)->(16 * i + j)>>
-///          to memref<?x?xf32>
-///   consumer %1 : memref<?x?xf32> ...
-/// ```
-///
-/// may fold into:
-///
-/// ```
-///   consumer %0 ... : memref<?x16xf32, affine_map<(i, j)->(16 * i + j)>>
-/// ```
-bool canFoldIntoConsumerOp(MemRefCastOp castOp);
-
-/// Compute `lhs` `pred` `rhs`, where `pred` is one of the known integer
-/// comparison predicates.
-bool applyCmpPredicate(CmpIPredicate predicate, const APInt &lhs,
-                       const APInt &rhs);
-
-/// Compute `lhs` `pred` `rhs`, where `pred` is one of the known floating point
-/// comparison predicates.
-bool applyCmpPredicate(CmpFPredicate predicate, const APFloat &lhs,
-                       const APFloat &rhs);
-
-/// Return true if ofr1 and ofr2 are the same integer constant attribute values
-/// or the same SSA value.
-/// Ignore integer bitwitdh and type mismatch that come from the fact there is
-/// no IndexAttr and that IndexType have no bitwidth.
-bool isEqualConstantIntOrValue(OpFoldResult ofr1, OpFoldResult ofr2);
-} // end namespace mlir
-
-#endif // MLIR_DIALECT_IR_STANDARDOPS_IR_OPS_H
+#endif // MLIR_DIALECT_MEMREF_IR_MEMREF_H_

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

+//===- MemRefBase.td - Base definitions for memref 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 MEMREF_BASE
+#define MEMREF_BASE
+
+include "mlir/IR/OpBase.td"
+
+def MemRef_Dialect : Dialect {
+  let name = "memref";
+  let cppNamespace = "::mlir::memref";
+  let description = [{
+    The `memref` dialect is intended to hold core memref creation and
+    manipulation ops, which are not strongly associated with any particular
+    other dialect or domain abstraction.
+  }];
+  let hasConstantMaterializer = 1;
+}
+
+#endif // MEMREF_BASE

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

+//===- MemRefOps.td - MemRef op definitions ----------------*- 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 MEMREF_OPS
+#define MEMREF_OPS
+
+include "mlir/Dialect/MemRef/IR/MemRefBase.td"
+include "mlir/Interfaces/CastInterfaces.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
+include "mlir/Interfaces/ViewLikeInterface.td"
+include "mlir/IR/SymbolInterfaces.td"
+
+class MemRef_Op<string mnemonic, list<OpTrait> traits = []>
+    : Op<MemRef_Dialect, mnemonic, traits> {
+  let printer = [{ return ::print(p, *this); }];
+  let verifier = [{ return ::verify(*this); }];
+  let parser = [{ return ::parse$cppClass(parser, result); }];
+}
+
+//===----------------------------------------------------------------------===//
+// AllocLikeOp
+//===----------------------------------------------------------------------===//
+
+// Base class for memref allocating ops: alloca and alloc.
+//
+//   %0 = alloclike(%m)[%s] : memref<8x?xf32, (d0, d1)[s0] -> ((d0 + s0), d1)>
+//
+class AllocLikeOp<string mnemonic,
+                  Resource resource,
+                  list<OpTrait> traits = []> :
+    MemRef_Op<mnemonic,
+    !listconcat([
+      AttrSizedOperandSegments
+    ], traits)> {
+
+  let arguments = (ins Variadic<Index>:$dynamicSizes,
+                   // The symbolic operands (the ones in square brackets) bind
+                   // to the symbols of the memref's layout map.
+                   Variadic<Index>:$symbolOperands,
+                   Confined<OptionalAttr<I64Attr>, [IntMinValue<0>]>:$alignment);
+  let results = (outs Res<AnyMemRef, "", [MemAlloc<resource>]>:$memref);
+
+  let builders = [
+    OpBuilder<(ins "MemRefType":$memrefType,
+                  CArg<"IntegerAttr", "IntegerAttr()">:$alignment), [{
+      return build($_builder, $_state, memrefType, {}, alignment);
+    }]>,
+    OpBuilder<(ins "MemRefType":$memrefType, "ValueRange":$dynamicSizes,
+                  CArg<"IntegerAttr", "IntegerAttr()">:$alignment), [{
+      return build($_builder, $_state, memrefType, dynamicSizes, {}, alignment);
+    }]>,
+    OpBuilder<(ins "MemRefType":$memrefType, "ValueRange":$dynamicSizes,
+                  "ValueRange":$symbolOperands,
+                  CArg<"IntegerAttr", "{}">:$alignment), [{
+      $_state.types.push_back(memrefType);
+      $_state.addOperands(dynamicSizes);
+      $_state.addOperands(symbolOperands);
+      $_state.addAttribute(getOperandSegmentSizeAttr(),
+          $_builder.getI32VectorAttr({
+              static_cast<int32_t>(dynamicSizes.size()),
+              static_cast<int32_t>(symbolOperands.size())}));
+      if (alignment)
+        $_state.addAttribute(getAlignmentAttrName(), alignment);
+    }]>];
+
+  let extraClassDeclaration = [{
+    static StringRef getAlignmentAttrName() { return "alignment"; }
+
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+
+    /// Returns the dynamic sizes for this alloc operation if specified.
+    operand_range getDynamicSizes() { return dynamicSizes(); }
+  }];
+
+  let assemblyFormat = [{
+    `(`$dynamicSizes`)` (`` `[` $symbolOperands^ `]`)? attr-dict `:` type($memref)
+  }];
+
+  let hasCanonicalizer = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// AssumeAlignmentOp
+//===----------------------------------------------------------------------===//
+
+def AssumeAlignmentOp : MemRef_Op<"assume_alignment"> {
+  let summary =
+      "assertion that gives alignment information to the input memref";
+  let description = [{
+    The `assume_alignment` operation takes a memref and an integer of alignment
+    value, and internally annotates the buffer with the given alignment. If
+    the buffer isn't aligned to the given alignment, the behavior is undefined.
+
+    This operation doesn't affect the semantics of a correct program. It's for
+    optimization only, and the optimization is best-effort.
+  }];
+  let arguments = (ins AnyMemRef:$memref,
+                       Confined<I32Attr, [IntPositive]>:$alignment);
+  let results = (outs);
+
+  let assemblyFormat = "$memref `,` $alignment attr-dict `:` type($memref)";
+}
+
+//===----------------------------------------------------------------------===//
+// BaseOpWithOffsetSizesAndStrides
+//===----------------------------------------------------------------------===//
+
+// Base class for ops with static/dynamic offset, sizes and strides
+// attributes/arguments.
+class BaseOpWithOffsetSizesAndStrides<string mnemonic, list<OpTrait> traits = []> :
+    MemRef_Op<mnemonic,
+           !listconcat(traits, [NoSideEffect, AttrSizedOperandSegments])> {
+  code extraBaseClassDeclaration = [{
+    /// Returns the dynamic sizes for this subview operation if specified.
+    operand_range getDynamicSizes() { return sizes(); }
+
+    /// Return the list of Range (i.e. offset, size, stride). Each
+    /// Range entry contains either the dynamic value or a ConstantIndexOp
+    /// constructed with `b` at location `loc`.
+    SmallVector<Range, 8> getOrCreateRanges(OpBuilder &b, Location loc) {
+      return mlir::getOrCreateRanges(*this, b, loc);
+    }
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// AllocOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_AllocOp : AllocLikeOp<"alloc", DefaultResource> {
+  let summary = "memory allocation operation";
+  let description = [{
+    The `alloc` operation allocates a region of memory, as specified by its
+    memref type.
+
+    Example:
+
+    ```mlir
+    %0 = memref.alloc() : memref<8x64xf32, 1>
+    ```
+
+    The optional list of dimension operands are bound to the dynamic dimensions
+    specified in its memref type. In the example below, the ssa value '%d' is
+    bound to the second dimension of the memref (which is dynamic).
+
+    ```mlir
+    %0 = memref.alloc(%d) : memref<8x?xf32, 1>
+    ```
+
+    The optional list of symbol operands are bound to the symbols of the
+    memrefs affine map. In the example below, the ssa value '%s' is bound to
+    the symbol 's0' in the affine map specified in the allocs memref type.
+
+    ```mlir
+    %0 = memref.alloc()[%s] : memref<8x64xf32,
+                              affine_map<(d0, d1)[s0] -> ((d0 + s0), d1)>, 1>
+    ```
+
+    This operation returns a single ssa value of memref type, which can be used
+    by subsequent load and store operations.
+
+    The optional `alignment` attribute may be specified to ensure that the
+    region of memory that will be indexed is aligned at the specified byte
+    boundary.
+
+    ```mlir
+    %0 = memref.alloc()[%s] {alignment = 8} :
+      memref<8x64xf32, affine_map<(d0, d1)[s0] -> ((d0 + s0), d1)>, 1>
+    ```
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// AllocaOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_AllocaOp : AllocLikeOp<"alloca", AutomaticAllocationScopeResource> {
+  let summary = "stack memory allocation operation";
+  let description = [{
+    The `alloca` operation allocates memory on the stack, to be automatically
+    released when control transfers back from the region of its closest
+    surrounding operation with an
+    [`AutomaticAllocationScope`](../Traits.md#automaticallocationscope) trait.
+    The amount of memory allocated is specified by its memref and additional
+    operands. For example:
+
+    ```mlir
+    %0 = memref.alloca() : memref<8x64xf32>
+    ```
+
+    The optional list of dimension operands are bound to the dynamic dimensions
+    specified in its memref type. In the example below, the SSA value '%d' is
+    bound to the second dimension of the memref (which is dynamic).
+
+    ```mlir
+    %0 = memref.alloca(%d) : memref<8x?xf32>
+    ```
+
+    The optional list of symbol operands are bound to the symbols of the
+    memref's affine map. In the example below, the SSA value '%s' is bound to
+    the symbol 's0' in the affine map specified in the allocs memref type.
+
+    ```mlir
+    %0 = memref.alloca()[%s] : memref<8x64xf32,
+                               affine_map<(d0, d1)[s0] -> ((d0 + s0), d1)>>
+    ```
+
+    This operation returns a single SSA value of memref type, which can be used
+    by subsequent load and store operations. An optional alignment attribute, if
+    specified, guarantees alignment at least to that boundary. If not specified,
+    an alignment on any convenient boundary compatible with the type will be
+    chosen.
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// BufferCastOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_BufferCastOp : MemRef_Op<"buffer_cast",
+    [SameOperandsAndResultShape, SameOperandsAndResultElementType,
+     TypesMatchWith<"type of 'tensor' is the tensor equivalent of 'memref'",
+                    "memref", "tensor",
+                    "getTensorTypeFromMemRefType($_self)">]> {
+  let summary = "tensor to memref cast operation";
+  let description = [{
+    Casts a tensor to a memref.
+
+    ```mlir
+    // Result type is tensor<4x?xf32>
+    %12 = memref.buffer_cast %10 : memref<4x?xf32, #map0, 42>
+    ```
+  }];
+
+  let arguments = (ins AnyTensor:$tensor);
+  let results = (outs AnyRankedOrUnrankedMemRef:$memref);
+  // This op is fully verified by traits.
+  let verifier = ?;
+
+  let assemblyFormat = "$tensor attr-dict `:` type($memref)";
+
+  let hasFolder = 1;
+  let hasCanonicalizer = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// CastOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_CastOp : MemRef_Op<"cast", [
+      NoSideEffect, SameOperandsAndResultShape,
+      DeclareOpInterfaceMethods<CastOpInterface>
+    ]> {
+  let summary = "memref cast operation";
+  let description = [{
+    Syntax:
+
+    ```
+    operation ::= ssa-id `=` `memref.cast` ssa-use `:` type `to` type
+    ```
+
+    The `memref.cast` operation converts a memref from one type to an equivalent
+    type with a compatible shape. The source and destination types are
+    compatible if:
+
+    a. Both are ranked memref types with the same element type, address space,
+    and rank and:
+      1. Both have the same layout or both have compatible strided layouts.
+      2. The individual sizes (resp. offset and strides in the case of strided
+         memrefs) may convert constant dimensions to dynamic dimensions and
+         vice-versa.
+
+    If the cast converts any dimensions from an unknown to a known size, then it
+    acts as an assertion that fails at runtime if the dynamic dimensions
+    disagree with resultant destination size.
+
+    Example:
+
+    ```mlir
+    // Assert that the input dynamic shape matches the destination static shape.
+    %2 = memref.cast %1 : memref<?x?xf32> to memref<4x4xf32>
+    // Erase static shape information, replacing it with dynamic information.
+    %3 = memref.cast %1 : memref<4xf32> to memref<?xf32>
+
+    // The same holds true for offsets and strides.
+
+    // Assert that the input dynamic shape matches the destination static stride.
+    %4 = memref.cast %1 : memref<12x4xf32, offset:?, strides: [?, ?]> to
+                          memref<12x4xf32, offset:5, strides: [4, 1]>
+    // Erase static offset and stride information, replacing it with
+    // dynamic information.
+    %5 = memref.cast %1 : memref<12x4xf32, offset:5, strides: [4, 1]> to
+                          memref<12x4xf32, offset:?, strides: [?, ?]>
+    ```
+
+    b. Either or both memref types are unranked with the same element type, and
+    address space.
+
+    Example:
+
+    ```mlir
+    Cast to concrete shape.
+        %4 = memref.cast %1 : memref<*xf32> to memref<4x?xf32>
+
+    Erase rank information.
+        %5 = memref.cast %1 : memref<4x?xf32> to memref<*xf32>
+    ```
+  }];
+
+  let arguments = (ins AnyRankedOrUnrankedMemRef:$source);
+  let results = (outs AnyRankedOrUnrankedMemRef:$dest);
+  let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
+  let verifier = "return impl::verifyCastOp(*this, areCastCompatible);";
+  let builders = [
+    OpBuilder<(ins "Value":$source, "Type":$destType), [{
+       impl::buildCastOp($_builder, $_state, source, destType);
+    }]>
+  ];
+
+  let extraClassDeclaration = [{
+    /// Fold the given CastOp into consumer op.
+    static bool canFoldIntoConsumerOp(CastOp castOp);
+  }];
+
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// DeallocOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_DeallocOp : MemRef_Op<"dealloc", [MemRefsNormalizable]> {
+  let summary = "memory deallocation operation";
+  let description = [{
+    The `dealloc` operation frees the region of memory referenced by a memref
+    which was originally created by the `alloc` operation.
+    The `dealloc` operation should not be called on memrefs which alias an
+    alloc'd memref (e.g. memrefs returned by `view` operations).
+
+    Example:
+
+    ```mlir
+    %0 = memref.alloc() : memref<8x64xf32, (d0, d1) -> (d0, d1), 1>
+    memref.dealloc %0 : memref<8x64xf32, (d0, d1) -> (d0, d1), 1>
+    ```
+  }];
+
+  let arguments = (ins Arg<AnyMemRef, "", [MemFree]>:$memref);
+
+  let hasCanonicalizer = 1;
+  let hasFolder = 1;
+  let assemblyFormat = "$memref attr-dict `:` type($memref)";
+}
+
+//===----------------------------------------------------------------------===//
+// DimOp
+//===----------------------------------------------------------------------===//
+
+def DimOp : MemRef_Op<"dim", [NoSideEffect]> {
+  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.
+
+    Example:
+
+    ```mlir
+    // Always returns 4, can be constant folded:
+    %c0 = constant 0 : index
+    %x = memref.dim %A, %c0 : memref<4 x ? x f32>
+
+    // Returns the dynamic dimension of %A.
+    %c1 = constant 1 : index
+    %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">:$memrefOrTensor,
+                       Index:$index);
+  let results = (outs Index:$result);
+
+  let assemblyFormat = [{
+    attr-dict $memrefOrTensor `,` $index `:` type($memrefOrTensor)
+  }];
+
+  let builders = [
+    OpBuilder<(ins "Value":$memrefOrTensor, "int64_t":$index)>,
+    OpBuilder<(ins "Value":$memrefOrTensor, "Value":$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;
+}
+
+//===----------------------------------------------------------------------===//
+// GetGlobalOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_GetGlobalOp : MemRef_Op<"get_global",
+    [NoSideEffect, DeclareOpInterfaceMethods<SymbolUserOpInterface>]> {
+  let summary = "get the memref pointing to a global variable";
+  let description = [{
+     The `memref.get_global` operation retrieves the memref pointing to a
+     named global variable. If the global variable is marked constant, writing
+     to the result memref (such as through a `memref.store` operation) is
+     undefined.
+
+     Example:
+
+     ```mlir
+     %x = memref.get_global @foo : memref<2xf32>
+     ```
+  }];
+
+  let arguments = (ins FlatSymbolRefAttr:$name);
+  let results = (outs AnyStaticShapeMemRef:$result);
+  let assemblyFormat = "$name `:` type($result) attr-dict";
+
+  // `GetGlobalOp` is fully verified by its traits.
+  let verifier = ?;
+}
+
+//===----------------------------------------------------------------------===//
+// GlobalOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_GlobalOp : MemRef_Op<"global", [Symbol]> {
+  let summary = "declare or define a global memref variable";
+  let description = [{
+    The `memref.global` operation declares or defines a named global variable.
+    The backing memory for the variable is allocated statically and is described
+    by the type of the variable (which should be a statically shaped memref
+    type). The operation is a declaration if no `inital_value` is specified,
+    else it is a definition. The `initial_value` can either be a unit attribute
+    to represent a definition of an uninitialized global variable, or an
+    elements attribute to represent the definition of a global variable with an
+    initial value. The global variable can also be marked constant using the
+    `constant` unit attribute. Writing to such constant global variables is
+    undefined.
+
+    The global variable can be accessed by using the `memref.get_global` to
+    retrieve the memref for the global variable. Note that the memref
+    for such global variable itself is immutable (i.e., memref.get_global for a
+    given global variable will always return the same memref descriptor).
+
+    Example:
+
+    ```mlir
+    // Private variable with an initial value.
+    memref.global "private" @x : memref<2xf32> = dense<0.0,2.0>
+
+    // Declaration of an external variable.
+    memref.global "private" @y : memref<4xi32>
+
+    // Uninitialized externally visible variable.
+    memref.global @z : memref<3xf16> = uninitialized
+
+    // Externally visible constant variable.
+    memref.global constant @c : memref<2xi32> = dense<1, 4>
+    ```
+  }];
+
+  let arguments = (ins
+      SymbolNameAttr:$sym_name,
+      OptionalAttr<StrAttr>:$sym_visibility,
+      TypeAttr:$type,
+      OptionalAttr<AnyAttr>:$initial_value,
+      UnitAttr:$constant
+  );
+
+  let assemblyFormat = [{
+       ($sym_visibility^)?
+       (`constant` $constant^)?
+       $sym_name `:`
+       custom<GlobalMemrefOpTypeAndInitialValue>($type, $initial_value)
+       attr-dict
+  }];
+
+  let extraClassDeclaration = [{
+     bool isExternal() { return !initial_value(); }
+     bool isUninitialized() {
+       return !isExternal() && initial_value().getValue().isa<UnitAttr>();
+     }
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// LoadOp
+//===----------------------------------------------------------------------===//
+
+def LoadOp : MemRef_Op<"load",
+     [TypesMatchWith<"result type matches element type of 'memref'",
+                     "memref", "result",
+                     "$_self.cast<MemRefType>().getElementType()">,
+                     MemRefsNormalizable]> {
+  let summary = "load operation";
+  let description = [{
+    The `load` op reads an element from a memref specified by an index list. The
+    output of load is a new value with the same type as the elements of the
+    memref. The arity of indices is the rank of the memref (i.e., if the memref
+    loaded from is of rank 3, then 3 indices are required for the load following
+    the memref identifier).
+
+    In an `affine.if` or `affine.for` body, the indices of a load are restricted
+    to SSA values bound to surrounding loop induction variables,
+    [symbols](Affine.md#dimensions-and-symbols), results of a
+    [`constant` operation](#stdconstant-constantop), or the result of an
+    `affine.apply` operation that can in turn take as arguments all of the
+    aforementioned SSA values or the recursively result of such an
+    `affine.apply` operation.
+
+    Example:
+
+    ```mlir
+    %1 = affine.apply affine_map<(d0, d1) -> (3*d0)> (%i, %j)
+    %2 = affine.apply affine_map<(d0, d1) -> (d1+1)> (%i, %j)
+    %12 = memref.load %A[%1, %2] : memref<8x?xi32, #layout, memspace0>
+
+    // Example of an indirect load (treated as non-affine)
+    %3 = affine.apply affine_map<(d0) -> (2*d0 + 1)>(%12)
+    %13 = memref.load %A[%3, %2] : memref<4x?xi32, #layout, memspace0>
+    ```
+
+    **Context:** The `load` and `store` operations are specifically crafted to
+    fully resolve a reference to an element of a memref, and (in affine
+    `affine.if` and `affine.for` operations) the compiler can follow use-def
+    chains (e.g. through [`affine.apply`](Affine.md#affineapply-affineapplyop)
+    operations) to precisely analyze references at compile-time using polyhedral
+    techniques. This is possible because of the
+    [restrictions on dimensions and symbols](Affine.md#restrictions-on-dimensions-and-symbols)
+    in these contexts.
+  }];
+
+  let arguments = (ins Arg<AnyMemRef, "the reference to load from",
+                           [MemRead]>:$memref,
+                       Variadic<Index>:$indices);
+  let results = (outs AnyType:$result);
+
+  let builders = [
+    OpBuilder<(ins "Value":$memref, CArg<"ValueRange", "{}">:$indices), [{
+      auto memrefType = memref.getType().cast<MemRefType>();
+      $_state.addOperands(memref);
+      $_state.addOperands(indices);
+      $_state.types.push_back(memrefType.getElementType());
+    }]>];
+
+  let extraClassDeclaration = [{
+    Value getMemRef() { return getOperand(0); }
+    void setMemRef(Value value) { setOperand(0, value); }
+    MemRefType getMemRefType() {
+      return getMemRef().getType().cast<MemRefType>();
+    }
+
+    operand_range getIndices() { return {operand_begin() + 1, operand_end()}; }
+  }];
+
+  let hasCanonicalizer = 1;
+  let hasFolder = 1;
+
+  let assemblyFormat = "$memref `[` $indices `]` attr-dict `:` type($memref)";
+}
+
+//===----------------------------------------------------------------------===//
+// PrefetchOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_PrefetchOp : MemRef_Op<"prefetch"> {
+  let summary = "prefetch operation";
+  let description = [{
+    The "prefetch" op prefetches data from a memref location described with
+    subscript indices similar to memref.load, and with three attributes: a
+    read/write specifier, a locality hint, and a cache type specifier as shown
+    below:
+
+    ```mlir
+    memref.prefetch %0[%i, %j], read, locality<3>, data : memref<400x400xi32>
+    ```
+
+    The read/write specifier is either 'read' or 'write', the locality hint
+    ranges from locality<0> (no locality) to locality<3> (extremely local keep
+    in cache). The cache type specifier is either 'data' or 'instr'
+    and specifies whether the prefetch is performed on data cache or on
+    instruction cache.
+  }];
+
+  let arguments = (ins AnyMemRef:$memref, Variadic<Index>:$indices,
+                   BoolAttr:$isWrite,
+                   Confined<I32Attr, [IntMinValue<0>,
+                     IntMaxValue<3>]>:$localityHint,
+                   BoolAttr:$isDataCache);
+
+  let extraClassDeclaration = [{
+    MemRefType getMemRefType() {
+      return memref().getType().cast<MemRefType>();
+    }
+    static StringRef getLocalityHintAttrName() { return "localityHint"; }
+    static StringRef getIsWriteAttrName() { return "isWrite"; }
+    static StringRef getIsDataCacheAttrName() { return "isDataCache"; }
+  }];
+
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// ReinterpretCastOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_ReinterpretCastOp:
+    BaseOpWithOffsetSizesAndStrides<"reinterpret_cast", [
+      NoSideEffect, ViewLikeOpInterface, OffsetSizeAndStrideOpInterface
+    ]> {
+  let summary = "memref reinterpret cast operation";
+  let description = [{
+    Modify offset, sizes and strides of an unranked/ranked memref.
+
+    Example:
+    ```mlir
+    memref.reinterpret_cast %ranked to
+      offset: [0],
+      sizes: [%size0, 10],
+      strides: [1, %stride1]
+    : memref<?x?xf32> to memref<?x10xf32, offset: 0, strides: [1, ?]>
+
+    memref.reinterpret_cast %unranked to
+      offset: [%offset],
+      sizes: [%size0, %size1],
+      strides: [%stride0, %stride1]
+    : memref<*xf32> to memref<?x?xf32, offset: ?, strides: [?, ?]>
+    ```
+  }];
+
+  let arguments = (ins
+    Arg<AnyRankedOrUnrankedMemRef, "", []>:$source,
+    Variadic<Index>:$offsets,
+    Variadic<Index>:$sizes,
+    Variadic<Index>:$strides,
+    I64ArrayAttr:$static_offsets,
+    I64ArrayAttr:$static_sizes,
+    I64ArrayAttr:$static_strides
+  );
+  let results = (outs AnyMemRef:$result);
+
+  let assemblyFormat = [{
+    $source `to` `offset` `` `:`
+    custom<OperandsOrIntegersOffsetsOrStridesList>($offsets, $static_offsets)
+    `` `,` `sizes` `` `:`
+    custom<OperandsOrIntegersSizesList>($sizes, $static_sizes) `` `,` `strides`
+    `` `:`
+    custom<OperandsOrIntegersOffsetsOrStridesList>($strides, $static_strides)
+    attr-dict `:` type($source) `to` type($result)
+  }];
+
+  let parser = ?;
+  let printer = ?;
+
+  let builders = [
+    // Build a ReinterpretCastOp with mixed static and dynamic entries.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "OpFoldResult":$offset, "ArrayRef<OpFoldResult>":$sizes,
+      "ArrayRef<OpFoldResult>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a ReinterpretCastOp with static entries.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "int64_t":$offset, "ArrayRef<int64_t>":$sizes,
+      "ArrayRef<int64_t>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a ReinterpretCastOp with dynamic entries.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "Value":$offset, "ValueRange":$sizes,
+      "ValueRange":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>
+  ];
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    // The result of the op is always a ranked memref.
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+    Value getViewSource() { return source(); }
+
+    /// Return the rank of the source ShapedType.
+    unsigned getResultRank() {
+      return getResult().getType().cast<ShapedType>().getRank();
+    }
+
+    /// Return the expected rank of each of the`static_offsets`, `static_sizes`
+    /// and `static_strides` attributes.
+    std::array<unsigned, 3> getArrayAttrMaxRanks() {
+      unsigned resultRank = getResult().getType().cast<ShapedType>().getRank();
+      return {1, resultRank, resultRank};
+    }
+
+    /// Return the number of leading operands before the `offsets`, `sizes` and
+    /// and `strides` operands.
+    static unsigned getOffsetSizeAndStrideStartOperandIndex() { return 1; }
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// ReshapeOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_ReshapeOp: MemRef_Op<"reshape", [
+    ViewLikeOpInterface, NoSideEffect]>  {
+  let summary = "memref reshape operation";
+  let description = [{
+    The `reshape` operation converts a memref from one type to an
+    equivalent type with a provided shape. The data is never copied or
+    modified. The source and destination types are compatible if both have the
+    same element type, same number of elements, address space and identity
+    layout map. The following combinations are possible:
+
+    a. Source type is ranked or unranked. Shape argument has static size.
+    Result type is ranked.
+
+    ```mlir
+    // Reshape statically-shaped memref.
+    %dst = memref.reshape %src(%shape)
+             : (memref<4x1xf32>, memref<1xi32>) to memref<4xf32>
+    %dst0 = memref.reshape %src(%shape0)
+             : (memref<4x1xf32>, memref<2xi32>) to memref<2x2xf32>
+    // Flatten unranked memref.
+    %dst = memref.reshape %src(%shape)
+             : (memref<*xf32>, memref<1xi32>) to memref<?xf32>
+    ```
+
+    b. Source type is ranked or unranked. Shape argument has dynamic size.
+    Result type is unranked.
+
+    ```mlir
+    // Reshape dynamically-shaped 1D memref.
+    %dst = memref.reshape %src(%shape)
+             : (memref<?xf32>, memref<?xi32>) to memref<*xf32>
+    // Reshape unranked memref.
+    %dst = memref.reshape %src(%shape)
+             : (memref<*xf32>, memref<?xi32>) to memref<*xf32>
+    ```
+  }];
+
+  let arguments = (ins
+    AnyRankedOrUnrankedMemRef:$source,
+    MemRefRankOf<[AnySignlessInteger, Index], [1]>:$shape
+  );
+  let results = (outs AnyRankedOrUnrankedMemRef:$result);
+
+  let builders = [OpBuilder<
+     (ins "MemRefType":$resultType, "Value":$operand, "Value":$shape), [{
+       $_state.addOperands(operand);
+       $_state.addOperands(shape);
+       $_state.addTypes(resultType);
+     }]>];
+
+  let extraClassDeclaration = [{
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+    Value getViewSource() { return source(); }
+  }];
+
+  let assemblyFormat = [{
+    $source `(` $shape `)` attr-dict `:` functional-type(operands, results)
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// StoreOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_StoreOp : MemRef_Op<"store",
+     [TypesMatchWith<"type of 'value' matches element type of 'memref'",
+                     "memref", "value",
+                     "$_self.cast<MemRefType>().getElementType()">,
+                     MemRefsNormalizable]> {
+  let summary = "store operation";
+  let description = [{
+    Store a value to a memref location given by indices. The value stored should
+    have the same type as the elemental type of the memref. The number of
+    arguments provided within brackets need to match the rank of the memref.
+
+    In an affine context, the indices of a store are restricted to SSA values
+    bound to surrounding loop induction variables,
+    [symbols](Affine.md#restrictions-on-dimensions-and-symbols), results of a
+    [`constant` operation](#stdconstant-constantop), or the result of an
+    [`affine.apply`](Affine.md#affineapply-affineapplyop) operation that can in
+    turn take as arguments all of the aforementioned SSA values or the
+    recursively result of such an `affine.apply` operation.
+
+    Example:
+
+    ```mlir
+    memref.store %100, %A[%1, 1023] : memref<4x?xf32, #layout, memspace0>
+    ```
+
+    **Context:** The `load` and `store` operations are specifically crafted to
+    fully resolve a reference to an element of a memref, and (in polyhedral
+    `affine.if` and `affine.for` operations) the compiler can follow use-def
+    chains (e.g. through [`affine.apply`](Affine.md#affineapply-affineapplyop)
+    operations) to precisely analyze references at compile-time using polyhedral
+    techniques. This is possible because of the
+    [restrictions on dimensions and symbols](Affine.md#restrictions-on-dimensions-and-symbols)
+    in these contexts.
+  }];
+
+  let arguments = (ins AnyType:$value,
+                       Arg<AnyMemRef, "the reference to store to",
+                           [MemWrite]>:$memref,
+                       Variadic<Index>:$indices);
+
+  let builders = [
+    OpBuilder<(ins "Value":$valueToStore, "Value":$memref), [{
+      $_state.addOperands(valueToStore);
+      $_state.addOperands(memref);
+    }]>];
+
+  let extraClassDeclaration = [{
+      Value getValueToStore() { return getOperand(0); }
+
+      Value getMemRef() { return getOperand(1); }
+      void setMemRef(Value value) { setOperand(1, value); }
+      MemRefType getMemRefType() {
+        return getMemRef().getType().cast<MemRefType>();
+      }
+
+      operand_range getIndices() {
+        return {operand_begin() + 2, operand_end()};
+      }
+  }];
+
+  let hasFolder = 1;
+
+  let assemblyFormat = [{
+    $value `,` $memref `[` $indices `]` attr-dict `:` type($memref)
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// SubViewOp
+//===----------------------------------------------------------------------===//
+
+def SubViewOp : BaseOpWithOffsetSizesAndStrides<
+    "subview", [DeclareOpInterfaceMethods<ViewLikeOpInterface>,
+                NoSideEffect, OffsetSizeAndStrideOpInterface] >  {
+  let summary = "memref subview operation";
+  let description = [{
+    The "subview" operation converts a memref type to another memref type
+    which represents a reduced-size view of the original memref as specified by
+    the operation's offsets, sizes and strides arguments.
+
+    The SubView operation supports the following arguments:
+
+    * source: the "base" memref on which to create a "view" memref.
+    * offsets: memref-rank number of offsets into the "base" memref at which to
+               create the "view" memref.
+    * sizes: memref-rank number of sizes which specify the sizes of the result
+             "view" memref type.
+    * strides: memref-rank number of strides that compose multiplicatively with
+               the base memref strides in each dimension.
+
+    The representation based on offsets, sizes and strides support a
+    partially-static specification via attributes specified through the
+    `static_offsets`, `static_sizes` and `static_strides` arguments. A special
+    sentinel value ShapedType::kDynamicSize and
+    ShapedType::kDynamicStrideOrOffset encodes that the corresponding entry has
+    a dynamic value.
+
+    A subview operation may additionally reduce the rank of the resulting view
+    by removing dimensions that are statically known to be of size 1.
+
+    Example 1:
+
+    ```mlir
+    %0 = memref.alloc() : memref<64x4xf32, (d0, d1) -> (d0 * 4 + d1)>
+
+    // Create a sub-view of "base" memref '%0' with offset arguments '%c0',
+    // dynamic sizes for each dimension, and stride arguments '%c1'.
+    %1 = memref.subview %0[%c0, %c0][%size0, %size1][%c1, %c1]
+      : memref<64x4xf32, (d0, d1) -> (d0 * 4 + d1) > to
+        memref<?x?xf32, (d0, d1)[s0, s1] -> (d0 * s1 + d1 + s0)>
+    ```
+
+    Example 2:
+
+    ```mlir
+    %0 = memref.alloc() : memref<8x16x4xf32, (d0, d1, d1) -> (d0 * 64 + d1 * 4 + d2)>
+
+    // Create a sub-view of "base" memref '%0' with dynamic offsets, sizes,
+    // and strides.
+    // Note that dynamic offsets are represented by the linearized dynamic
+    // offset symbol 's0' in the subview memref layout map, and that the
+    // dynamic strides operands, after being applied to the base memref
+    // strides in each dimension, are represented in the view memref layout
+    // map as symbols 's1', 's2' and 's3'.
+    %1 = memref.subview %0[%i, %j, %k][%size0, %size1, %size2][%x, %y, %z]
+      : memref<8x16x4xf32, (d0, d1, d2) -> (d0 * 64 + d1 * 4 + d2)> to
+        memref<?x?x?xf32,
+          (d0, d1, d2)[s0, s1, s2, s3] -> (d0 * s1 + d1 * s2 + d2 * s3 + s0)>
+    ```
+
+    Example 3:
+
+    ```mlir
+    %0 = memref.alloc() : memref<8x16x4xf32, (d0, d1, d1) -> (d0 * 64 + d1 * 4 + d2)>
+
+    // Subview with constant offsets, sizes and strides.
+    %1 = memref.subview %0[0, 2, 0][4, 4, 4][64, 4, 1]
+      : memref<8x16x4xf32, (d0, d1, d2) -> (d0 * 64 + d1 * 4 + d2)> to
+        memref<4x4x4xf32, (d0, d1, d2) -> (d0 * 64 + d1 * 4 + d2 + 8)>
+    ```
+
+    Example 4:
+
+    ```mlir
+    %0 = memref.alloc(%arg0, %arg1) : memref<?x?xf32>
+
+    // Subview with constant size, but dynamic offsets and
+    // strides. The resulting memref has a static shape, but if the
+    // base memref has an affine map to describe the layout, the result
+    // memref also uses an affine map to describe the layout. The
+    // strides of the result memref is computed as follows:
+    //
+    // Let #map1 represents the layout of the base memref, and #map2
+    // represents the layout of the result memref. A #mapsubview can be
+    // constructed to map an index from the result memref to the base
+    // memref (note that the description below uses more convenient
+    // naming for symbols, while in affine maps, symbols are
+    // represented as unsigned numbers that identify that symbol in the
+    // given affine map.
+    //
+    // #mapsubview = (d0, d1)[o0, o1, t0, t1] -> (d0 * t0 + o0, d1 * t1 + o1)
+    //
+    // where, o0, o1, ... are offsets, and t0, t1, ... are strides. Then,
+    //
+    // #map2 = #map1.compose(#mapsubview)
+    //
+    // If the layout map is represented as
+    //
+    // #map1 = (d0, d1)[s0, s1, s2] -> (d0 * s1 + d1 * s2 + s0)
+    //
+    // then,
+    //
+    // #map2 = (d0, d1)[s0, s1, s2, o0, o1, t0, t1] ->
+    //              (d0 * s1 * t0 + d1 * s2 * t1 + o0 * s1 + o1 * s2 + s0)
+    //
+    // Representing this canonically
+    //
+    // #map2 = (d0, d1)[r0, r1, r2] -> (d0 * r1 + d1 * r2 + r0)
+    //
+    // where, r0 = o0 * s1 + o1 * s2 + s0, r1 = s1 * t0, r2 = s2 * t1.
+    %1 = memref.subview %0[%i, %j][4, 4][%x, %y] :
+      : memref<?x?xf32, (d0, d1)[s0, s1, s2] -> (d0 * s1 + d1 * s2 + s0)> to
+        memref<4x4xf32, (d0, d1)[r0, r1, r2] -> (d0 * r1 + d1 * r2 + r0)>
+
+    // Note that the subview op does not guarantee that the result
+    // memref is "inbounds" w.r.t to base memref. It is upto the client
+    // to ensure that the subview is accessed in a manner that is
+    // in-bounds.
+    ```
+
+    Example 5:
+
+    ```mlir
+    // Rank-reducing subview.
+    %1 = memref.subview %0[0, 0, 0][1, 16, 4][1, 1, 1] :
+      memref<8x16x4xf32> to memref<16x4xf32>
+    %3 = memref.subview %2[3, 4, 2][1, 6, 3][1, 1, 1] :
+      memref<8x16x4xf32> to memref<6x3xf32, offset: 210, strides: [4, 1]>
+    ```
+    }
+  }];
+
+  let arguments = (ins
+    AnyMemRef:$source,
+    Variadic<Index>:$offsets,
+    Variadic<Index>:$sizes,
+    Variadic<Index>:$strides,
+    I64ArrayAttr:$static_offsets,
+    I64ArrayAttr:$static_sizes,
+    I64ArrayAttr:$static_strides
+  );
+  let results = (outs AnyMemRef:$result);
+
+  let assemblyFormat = [{
+    $source ``
+    custom<OperandsOrIntegersOffsetsOrStridesList>($offsets, $static_offsets)
+    custom<OperandsOrIntegersSizesList>($sizes, $static_sizes)
+    custom<OperandsOrIntegersOffsetsOrStridesList>($strides, $static_strides)
+    attr-dict `:` type($source) `to` type($result)
+  }];
+
+  let builders = [
+    // Build a SubViewOp with mixed static and dynamic entries and custom
+    // result type. If the type passed is nullptr, it is inferred.
+    OpBuilder<(ins "Value":$source, "ArrayRef<OpFoldResult>":$offsets,
+      "ArrayRef<OpFoldResult>":$sizes, "ArrayRef<OpFoldResult>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a SubViewOp with mixed static and dynamic entries and inferred
+    // result type.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "ArrayRef<OpFoldResult>":$offsets, "ArrayRef<OpFoldResult>":$sizes,
+      "ArrayRef<OpFoldResult>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a SubViewOp with static entries and custom result type. If the
+    // type passed is nullptr, it is inferred.
+    OpBuilder<(ins "Value":$source, "ArrayRef<int64_t>":$offsets,
+      "ArrayRef<int64_t>":$sizes, "ArrayRef<int64_t>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a SubViewOp with static entries and inferred result type.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "ArrayRef<int64_t>":$offsets, "ArrayRef<int64_t>":$sizes,
+      "ArrayRef<int64_t>":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a SubViewOp with dynamic entries and custom result type. If the
+    // type passed is nullptr, it is inferred.
+    OpBuilder<(ins "Value":$source, "ValueRange":$offsets,
+      "ValueRange":$sizes, "ValueRange":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
+    // Build a SubViewOp with dynamic entries and inferred result type.
+    OpBuilder<(ins "MemRefType":$resultType, "Value":$source,
+      "ValueRange":$offsets, "ValueRange":$sizes, "ValueRange":$strides,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>
+  ];
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    /// Returns the type of the base memref operand.
+    MemRefType getSourceType() {
+      return source().getType().cast<MemRefType>();
+    }
+
+    /// The result of a subview is always a memref.
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+
+    /// A subview result type can be fully inferred from the source type and the
+    /// static representation of offsets, sizes and strides. Special sentinels
+    /// encode the dynamic case.
+    static Type inferResultType(MemRefType sourceMemRefType,
+                                ArrayRef<int64_t> staticOffsets,
+                                ArrayRef<int64_t> staticSizes,
+                                ArrayRef<int64_t> staticStrides);
+    static Type inferResultType(MemRefType sourceMemRefType,
+                                ArrayRef<OpFoldResult> staticOffsets,
+                                ArrayRef<OpFoldResult> staticSizes,
+                                ArrayRef<OpFoldResult> staticStrides);
+    static Type inferRankReducedResultType(unsigned resultRank,
+                                           MemRefType sourceMemRefType,
+                                           ArrayRef<int64_t> staticOffsets,
+                                           ArrayRef<int64_t> staticSizes,
+                                           ArrayRef<int64_t> staticStrides);
+    static Type inferRankReducedResultType(unsigned resultRank,
+                                           MemRefType sourceMemRefType,
+                                           ArrayRef<OpFoldResult> staticOffsets,
+                                           ArrayRef<OpFoldResult> staticSizes,
+                                           ArrayRef<OpFoldResult> staticStrides);
+
+    /// Return the expected rank of each of the`static_offsets`, `static_sizes`
+    /// and `static_strides` attributes.
+    std::array<unsigned, 3> getArrayAttrMaxRanks() {
+      unsigned rank = getSourceType().getRank();
+      return {rank, rank, rank};
+    }
+
+    /// Return the number of leading operands before the `offsets`, `sizes` and
+    /// and `strides` operands.
+    static unsigned getOffsetSizeAndStrideStartOperandIndex() { return 1; }
+  }];
+
+  let hasCanonicalizer = 1;
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// TensorLoadOp
+//===----------------------------------------------------------------------===//
+
+def TensorLoadOp : MemRef_Op<"tensor_load",
+    [SameOperandsAndResultShape, SameOperandsAndResultElementType,
+     TypesMatchWith<"result type matches tensor equivalent of 'memref'",
+                    "memref", "result",
+                    "getTensorTypeFromMemRefType($_self)">]> {
+  let summary = "tensor load operation";
+  let description = [{
+    Create a tensor from a memref, making an independent copy of the element
+    data. The result value is a tensor whose shape and element type match the
+    memref operand.
+
+    The opposite of this op is buffer_cast. Together, these two ops are
+    useful for source/target materializations when doing type conversions
+    involving tensors and memrefs.
+
+    Example:
+
+    ```mlir
+    // Produces a value of tensor<4x?xf32> type.
+    %12 = memref.tensor_load %10 : memref<4x?xf32, #layout, memspace0>
+    ```
+  }];
+
+  let arguments = (ins Arg<AnyRankedOrUnrankedMemRef,
+                       "the reference to load from", [MemRead]>:$memref);
+  let results = (outs AnyTensor:$result);
+  // TensorLoadOp is fully verified by traits.
+  let verifier = ?;
+
+  let builders = [
+    OpBuilder<(ins "Value":$memref), [{
+      $_state.addOperands(memref);
+      $_state.addTypes(getTensorTypeFromMemRefType(memref.getType()));
+    }]>];
+
+  let extraClassDeclaration = [{
+    /// The result of a tensor_load is always a tensor.
+    TensorType getType() {
+      Type resultType = getResult().getType();
+      if (resultType.isa<TensorType>())
+        return resultType.cast<TensorType>();
+      return {};
+    }
+  }];
+
+  let assemblyFormat = "$memref attr-dict `:` type($memref)";
+
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// TensorStoreOp
+//===----------------------------------------------------------------------===//
+
+def TensorStoreOp : MemRef_Op<"tensor_store",
+    [SameOperandsShape, SameOperandsElementType,
+     TypesMatchWith<"type of 'value' matches tensor equivalent of 'memref'",
+                    "memref", "tensor",
+                    "getTensorTypeFromMemRefType($_self)">]> {
+  let summary = "tensor store operation";
+  let description = [{
+    Stores the contents of a tensor into a memref. The first operand is a value
+    of tensor type, the second operand is a value of memref type. The shapes and
+    element types of these must match, and are specified by the memref type.
+
+    Example:
+
+    ```mlir
+    %9 = dim %8, 1 : tensor<4x?xf32>
+    %10 = alloc(%9) : memref<4x?xf32, #layout, memspace0>
+    memref.tensor_store %8, %10 : memref<4x?xf32, #layout, memspace0>
+    ```
+  }];
+
+  let arguments = (ins AnyTensor:$tensor, Arg<AnyRankedOrUnrankedMemRef,
+                       "the reference to store to", [MemWrite]>:$memref);
+  // TensorStoreOp is fully verified by traits.
+  let verifier = ?;
+
+  let assemblyFormat = "$tensor `,` $memref attr-dict `:` type($memref)";
+}
+
+//===----------------------------------------------------------------------===//
+// TransposeOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_TransposeOp : MemRef_Op<"transpose", [NoSideEffect]>,
+    Arguments<(ins AnyStridedMemRef:$in, AffineMapAttr:$permutation)>,
+    Results<(outs AnyStridedMemRef)> {
+  let summary = "`transpose` produces a new strided memref (metadata-only)";
+  let description = [{
+    The `transpose` op produces a strided memref whose sizes and strides
+    are a permutation of the original `in` memref. This is purely a metadata
+    transformation.
+
+    Example:
+
+    ```mlir
+    %1 = memref.transpose %0 (i, j) -> (j, i) : memref<?x?xf32> to memref<?x?xf32, affine_map<(d0, d1)[s0] -> (d1 * s0 + d0)>>
+    ```
+  }];
+
+  let builders = [
+    OpBuilder<(ins "Value":$in, "AffineMapAttr":$permutation,
+      CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>];
+
+  let extraClassDeclaration = [{
+    static StringRef getPermutationAttrName() { return "permutation"; }
+    ShapedType getShapedType() { return in().getType().cast<ShapedType>(); }
+  }];
+
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
+// ViewOp
+//===----------------------------------------------------------------------===//
+
+def MemRef_ViewOp : MemRef_Op<"view", [
+    DeclareOpInterfaceMethods<ViewLikeOpInterface>, NoSideEffect]> {
+  let summary = "memref view operation";
+  let description = [{
+    The "view" operation extracts an N-D contiguous memref with empty layout map
+    with arbitrary element type from a 1-D contiguous memref with empty layout
+    map of i8 element  type. The ViewOp supports the following arguments:
+
+    * A single dynamic byte-shift operand must be specified which represents a
+      a shift of the base 1-D memref pointer from which to create the resulting
+      contiguous memref view with identity layout.
+    * A dynamic size operand that must be specified for each dynamic dimension
+      in the resulting view memref type.
+
+    The "view" operation gives a structured indexing form to a flat 1-D buffer.
+    Unlike "subview" it can perform a type change. The type change behavior
+    requires the op to have special semantics because, e.g. a byte shift of 3
+    cannot be represented as an offset on f64.
+    For now, a "view" op:
+
+    1. Only takes a contiguous source memref with 0 offset and empty layout.
+    2. Must specify a byte_shift operand (in the future, a special integer
+       attribute may be added to support the folded case).
+    3. Returns a contiguous memref with 0 offset and empty layout.
+
+    Example:
+
+    ```mlir
+    // Allocate a flat 1D/i8 memref.
+    %0 = memref.alloc() : memref<2048xi8>
+
+    // ViewOp with dynamic offset and static sizes.
+    %1 = memref.view %0[%offset_1024][] : memref<2048xi8> to memref<64x4xf32>
+
+    // ViewOp with dynamic offset and two dynamic size.
+    %2 = memref.view %0[%offset_1024][%size0, %size1] :
+      memref<2048xi8> to memref<?x4x?xf32>
+    ```
+  }];
+
+  let arguments = (ins MemRefRankOf<[I8], [1]>:$source,
+                   Index:$byte_shift,
+                   Variadic<Index>:$sizes);
+  let results = (outs AnyMemRef);
+
+  let extraClassDeclaration = [{
+    /// The result of a view is always a memref.
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+
+    /// Returns the dynamic sizes for this view operation. This is redundant
+    /// with `sizes` but needed in template implementations. More specifically:
+    /// ```
+    /// template <typename AnyMemRefDefOp>
+    /// bool isMemRefSizeValidSymbol(AnyMemRefDefOp memrefDefOp, unsigned index,
+    ///                              Region *region)
+    /// ```
+    operand_range getDynamicSizes() {
+      return {sizes().begin(), sizes().end()};
+    }
+  }];
+
+  let hasCanonicalizer = 1;
+}
+
+#endif // MEMREF_OPS

mlir/include/mlir/Dialect/StandardOps/Utils/Utils.h

 // either by passes, optimization sequences, or in turn by other transformation
 // utilities.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef MLIR_DIALECT_STANDARDOPS_UTILS_UTILS_H
 #define MLIR_DIALECT_STANDARDOPS_UTILS_UTILS_H
 
+#include "mlir/Dialect/StandardOps/IR/Ops.h"
+#include "mlir/IR/Matchers.h"
+#include "mlir/IR/PatternMatch.h"
 #include "mlir/IR/Value.h"
 
 namespace mlir {
 
 class Location;
 class OpBuilder;
 
 /// 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);
 
+/// Matches a ConstantIndexOp.
+detail::op_matcher<ConstantIndexOp> matchConstantIndex();
+
+/// Detects the `values` produced by a ConstantIndexOp and places the new
+/// constant in place of the corresponding sentinel value.
+void canonicalizeSubViewPart(SmallVectorImpl<OpFoldResult> &values,
+                             function_ref<bool(int64_t)> isDynamic);
+
+void getPositionsOfShapeOne(unsigned rank, ArrayRef<int64_t> shape,
+                            llvm::SmallDenseSet<unsigned> &dimsToProject);
+
+/// Pattern to rewrite a subview op with constant arguments.
+template <typename OpType, typename CastOpFunc>
+class OpWithOffsetSizesAndStridesConstantArgumentFolder final
+    : public OpRewritePattern<OpType> {
+public:
+  using OpRewritePattern<OpType>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(OpType op,
+                                PatternRewriter &rewriter) const override {
+    // No constant operand, just return;
+    if (llvm::none_of(op.getOperands(), [](Value operand) {
+          return matchPattern(operand, matchConstantIndex());
+        }))
+      return failure();
+
+    // At least one of offsets/sizes/strides is a new constant.
+    // Form the new list of operands and constant attributes from the existing.
+    SmallVector<OpFoldResult> mixedOffsets(op.getMixedOffsets());
+    SmallVector<OpFoldResult> mixedSizes(op.getMixedSizes());
+    SmallVector<OpFoldResult> mixedStrides(op.getMixedStrides());
+    canonicalizeSubViewPart(mixedOffsets, ShapedType::isDynamicStrideOrOffset);
+    canonicalizeSubViewPart(mixedSizes, ShapedType::isDynamic);
+    canonicalizeSubViewPart(mixedStrides, ShapedType::isDynamicStrideOrOffset);
+
+    // Create the new op in canonical form.
+    auto newOp = rewriter.create<OpType>(op.getLoc(), op.source(), mixedOffsets,
+                                         mixedSizes, mixedStrides);
+    CastOpFunc func;
+    func(rewriter, op, newOp);
+
+    return success();
+  }
+};
+
 } // end namespace mlir
 
 #endif // MLIR_DIALECT_STANDARDOPS_UTILS_UTILS_H

mlir/include/mlir/Transforms/Passes.td

let description = [{
 
     All memref-typed results are appended to the function argument list.
 
     The main issue with this pass (and the out-param calling convention) is that
     buffers for results need to be allocated in the caller. This currently only
     works for static shaped memrefs.
   }];
   let constructor = "mlir::createBufferResultsToOutParamsPass()";
-  let dependentDialects = ["linalg::LinalgDialect"];
+  let dependentDialects = ["linalg::LinalgDialect", "memref::MemRefDialect"];
 }
 
 def Canonicalizer : Pass<"canonicalize"> {
   let summary = "Canonicalize operations";
   let description = [{
     This pass performs various types of canonicalizations over a set of
     operations. See [Operation Canonicalization](Canonicalization.md) for more
     details.
   }];
   let constructor = "mlir::createCanonicalizerPass()";
+  let dependentDialects = ["memref::MemRefDialect"];
 }
 
 def CopyRemoval : FunctionPass<"copy-removal"> {
   let summary = "Remove the redundant copies from input IR";
   let constructor = "mlir::createCopyRemovalPass()";
 }
 
 def CSE : Pass<"cse"> {
Option<"maxInliningIterations", "max-iterations", "unsigned",
            "Maximum number of iterations when inlining within an SCC">,
   ];
 }
 
 def FinalizingBufferize : FunctionPass<"finalizing-bufferize"> {
   let summary = "Finalize a partial bufferization";
   let description = [{
     A bufferize pass that finalizes a partial bufferization by removing
-    remaining `tensor_load` and `tensor_to_memref` operations.
+    remaining `memref.tensor_load` and `memref.buffer_cast` operations.
 
     The removal of those operations is only possible if the operations only
-    exist in pairs, i.e., all uses of `tensor_load` operations are
-    `tensor_to_memref` operations.
+    exist in pairs, i.e., all uses of `memref.tensor_load` operations are
+    `memref.buffer_cast` operations.
 
     This pass will fail if not all operations can be removed or if any operation
     with tensor typed operands remains.
   }];
   let constructor = "mlir::createFinalizingBufferizePass()";
 }
 
 def LocationSnapshot : Pass<"snapshot-op-locations"> {
let description = [{
     memref types while preserving the original behavior, users of those
     memref types are also modified to incorporate the resulting layout map.
     For instance, an [AffineLoadOp]
     (https://mlir.llvm.org/docs/Dialects/Affine/#affineload-affineloadop)
     will be updated to compose the layout map with with the affine expression
     contained in the op. Operations marked with the [MemRefsNormalizable]
     (https://mlir.llvm.org/docs/Traits/#memrefsnormalizable) trait are
     expected to be normalizable. Supported operations include affine
-    operations, std.alloc, std.dealloc, and std.return.
+    operations, memref.alloc, memref.dealloc, and std.return.
 
     Given an appropriate layout map specified in the code, this transformation
     can express tiled or linearized access to multi-dimensional data
     structures, but will not modify memref types without an explicit layout
     map.
 
     Currently this pass is limited to only modify
     functions where all memref types can be normalized. If a function

mlir/include/mlir/Transforms/Utils.h

 
 namespace mlir {
 
 class AffineApplyOp;
 class AffineForOp;
 class Location;
 class OpBuilder;
 
+namespace memref {
+class AllocOp;
+} // end namespace memref
+
 /// Replaces all "dereferencing" uses of `oldMemRef` with `newMemRef` while
 /// optionally remapping the old memref's indices using the supplied affine map,
 /// `indexRemap`. The new memref could be of a different shape or rank.
 /// `extraIndices` provides any additional access indices to be added to the
 /// start.
 ///
 /// `indexRemap` remaps indices of the old memref access to a new set of indices
 /// that are used to index the memref. Additional input operands to indexRemap
LogicalResult replaceAllMemRefUsesWith(Value oldMemRef, Value newMemRef,
                                        AffineMap indexRemap = AffineMap(),
                                        ArrayRef<Value> extraOperands = {},
                                        ArrayRef<Value> symbolOperands = {},
                                        bool allowNonDereferencingOps = false);
 
 /// Rewrites the memref defined by this alloc op to have an identity layout map
 /// and updates all its indexing uses. Returns failure if any of its uses
 /// escape (while leaving the IR in a valid state).
-LogicalResult normalizeMemRef(AllocOp op);
+LogicalResult normalizeMemRef(memref::AllocOp *op);
 
 /// Uses the old memref type map layout and computes the new memref type to have
 /// a new shape and a layout map, where the old layout map has been normalized
 /// to an identity layout map. It returns the old memref in case no
 /// normalization was needed or a failure occurs while transforming the old map
 /// layout to an identity layout map.
 MemRefType normalizeMemRefType(MemRefType memrefType, OpBuilder builder,
                                unsigned numSymbolicOperands);

mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp

 
 #include "mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h"
 
 #include "mlir/Conversion/StandardToLLVM/ConvertStandardToLLVMPass.h"
 #include "mlir/Dialect/GPU/GPUDialect.h"
 #include "mlir/Dialect/GPU/Passes.h"
 #include "mlir/Dialect/LLVMIR/NVVMDialect.h"
 #include "mlir/Dialect/Math/IR/Math.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/IR/BlockAndValueMapping.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "llvm/Support/FormatVariadic.h"
 
 #include "../GPUCommon/GPUOpsLowering.h"
 #include "../GPUCommon/IndexIntrinsicsOpLowering.h"
 #include "../GPUCommon/OpToFuncCallLowering.h"

mlir/lib/Dialect/Affine/Transforms/AffineDataCopyGeneration.cpp

 // performed on already tiled code.
 //
 //===----------------------------------------------------------------------===//
 
 #include "PassDetail.h"
 #include "mlir/Analysis/Utils.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Affine/Passes.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "mlir/Transforms/LoopUtils.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include <algorithm>
 

mlir/lib/Dialect/GPU/Transforms/MemoryPromotion.cpp

 //===- MemoryPromotion.cpp - Utilities for moving data across GPU memories ===//
 //
 // 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 file implements utilities that allow one to create IR moving the data
 // across different levels of the GPU memory hierarchy.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/GPU/MemoryPromotion.h"
 #include "mlir/Dialect/GPU/GPUDialect.h"
+#include "mlir/Dialect/MemRef/EDSC/Intrinsics.h"
 #include "mlir/Dialect/SCF/EDSC/Builders.h"
 #include "mlir/Dialect/StandardOps/EDSC/Intrinsics.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/LoopUtils.h"
 
 using namespace mlir;
 using namespace mlir::edsc;
 using namespace mlir::edsc::intrinsics;
blockDims.push_back(
         builder.create<gpu::BlockDimOp>(loc, indexType, dimName));
   }
 
   // Produce the loop nest with copies.
   SmallVector<Value, 8> ivs(lbs.size());
   loopNestBuilder(lbs, ubs, steps, [&](ValueRange loopIvs) {
     ivs.assign(loopIvs.begin(), loopIvs.end());
     auto activeIvs = llvm::makeArrayRef(ivs).take_back(rank);
-    StdIndexedValue fromHandle(from), toHandle(to);
+    MemRefIndexedValue fromHandle(from), toHandle(to);
     toHandle(activeIvs) = fromHandle(activeIvs);
   });
 
   // Map the innermost loops to threads in reverse order.
   for (auto en :
        llvm::enumerate(llvm::reverse(llvm::makeArrayRef(ivs).take_back(
            GPUDialect::getNumWorkgroupDimensions())))) {
     Value v = en.value();

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

 //===- LinalgInterfaces.cpp - Linalg interfaces implementation ------------===//
 //
 // 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/Linalg/IR/LinalgInterfaces.h"
 
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/IR/AffineExprVisitor.h"
 #include "mlir/IR/AffineMap.h"
 #include "llvm/ADT/SmallSet.h"
 
 using namespace mlir;
 using namespace mlir::linalg;
 
 /// Include the definitions of the copy operation interface.

 }
 
 SmallVector<Value, 4> LinalgOp::createFlatListOfOperandDims(OpBuilder &b,
                                                             Location loc) {
   SmallVector<Value, 4> res;
   for (Value v : getShapedOperands()) {
     ShapedType t = v.getType().template cast<ShapedType>();
     for (unsigned i = 0, e = t.getRank(); i < e; ++i)
-      res.push_back(b.create<DimOp>(loc, v, i));
+      res.push_back(b.create<memref::DimOp>(loc, v, i));
   }
   return res;
 }
 
 SmallVector<Range, 4> LinalgOp::createLoopRanges(OpBuilder &b, Location loc) {
   AffineMap map = getLoopsToShapesMap();
   unsigned numDims = map.getNumDims(), numRes = map.getNumResults();
   auto viewSizes = createFlatListOfOperandDims(b, loc);

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

 //===- Bufferize.cpp - scf bufferize pass ---------------------------------===//
 //
 // 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/Transforms/Bufferize.h"
 #include "PassDetail.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/SCF/Passes.h"
 #include "mlir/Dialect/SCF/SCF.h"
 #include "mlir/Dialect/SCF/Transforms.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 using namespace mlir;
 using namespace mlir::scf;
 
 namespace {
 struct SCFBufferizePass : public SCFBufferizeBase<SCFBufferizePass> {
   void runOnFunction() override {
     auto func = getOperation();
     auto *context = &getContext();
 
     BufferizeTypeConverter typeConverter;
     OwningRewritePatternList patterns;
     ConversionTarget target(*context);
 
     populateBufferizeMaterializationLegality(target);
     populateSCFStructuralTypeConversionsAndLegality(context, typeConverter,

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

 //====----- Bufferize.cpp - Bufferization of shape ops  ---------*- 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
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Transforms/Bufferize.h"
 #include "PassDetail.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/Shape/Transforms/Passes.h"
 #include "mlir/Pass/Pass.h"
 
 using namespace mlir;
 
 namespace {
 struct ShapeBufferizePass : public ShapeBufferizeBase<ShapeBufferizePass> {
   void runOnFunction() override {
     MLIRContext &ctx = getContext();
 
     OwningRewritePatternList patterns;
     BufferizeTypeConverter typeConverter;
     ConversionTarget target(getContext());
 
     populateBufferizeMaterializationLegality(target);
     populateShapeStructuralTypeConversionsAndLegality(&ctx, typeConverter,
                                                       patterns, target);

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

 //===- Bufferize.cpp - Bufferization for std ops --------------------------===//
 //
 // 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 file implements bufferization of std.func's and std.call's.
 //
 //===----------------------------------------------------------------------===//
 
 #include "PassDetail.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/StandardOps/Transforms/Passes.h"
 #include "mlir/Transforms/Bufferize.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 using namespace mlir;
 
 namespace {
 struct FuncBufferizePass : public FuncBufferizeBase<FuncBufferizePass> {
   using FuncBufferizeBase<FuncBufferizePass>::FuncBufferizeBase;
-
   void runOnOperation() override {
     auto module = getOperation();
     auto *context = &getContext();
 
     BufferizeTypeConverter typeConverter;
     OwningRewritePatternList patterns;
     ConversionTarget target(*context);
 
     populateFuncOpTypeConversionPattern(patterns, context, typeConverter);
     target.addDynamicallyLegalOp<FuncOp>([&](FuncOp op) {
       return typeConverter.isSignatureLegal(op.getType()) &&
              typeConverter.isLegal(&op.getBody());
     });
     populateCallOpTypeConversionPattern(patterns, context, typeConverter);
     target.addDynamicallyLegalOp<CallOp>(
         [&](CallOp op) { return typeConverter.isLegal(op); });
 
     populateBranchOpInterfaceTypeConversionPattern(patterns, context,
                                                    typeConverter);
     populateReturnOpTypeConversionPattern(patterns, context, typeConverter);
-    target.addLegalOp<ModuleOp, ModuleTerminatorOp, TensorLoadOp,
-                      TensorToMemrefOp>();
+    target.addLegalOp<ModuleOp, ModuleTerminatorOp, memref::TensorLoadOp,
+                      memref::BufferCastOp>();
 
     target.markUnknownOpDynamicallyLegal([&](Operation *op) {
       return isNotBranchOpInterfaceOrReturnLikeOp(op) ||
              isLegalForBranchOpInterfaceTypeConversionPattern(op,
                                                               typeConverter) ||
              isLegalForReturnOpTypeConversionPattern(op, typeConverter);
     });
 

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

 //===- Bufferize.cpp - Bufferization for std ops --------------------------===//
 //
 // 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 file implements bufferization of tensor-valued std.constant ops.
 //
 //===----------------------------------------------------------------------===//
 
 #include "PassDetail.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/StandardOps/Transforms/Passes.h"
 #include "mlir/IR/BlockAndValueMapping.h"
 #include "mlir/Transforms/Bufferize.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 using namespace mlir;
 
 namespace {
 // This class creates global ops for all tensor-valued constants in the program.
 // It creates them with pretty names and makes sure that duplicate globals
 // aren't created.
 class GlobalCreator {
 public:
   explicit GlobalCreator(ModuleOp module);
-  GlobalMemrefOp getGlobalFor(Attribute attr) {
+  memref::GlobalOp getGlobalFor(Attribute attr) {
     assert(globals.find(attr) != globals.end() && "unknown constant attr");
     return globals[attr];
   }
 
 private:
-  DenseMap<Attribute, GlobalMemrefOp> globals;
+  DenseMap<Attribute, memref::GlobalOp> globals;
 };
 
 GlobalCreator::GlobalCreator(ModuleOp module) {
   BufferizeTypeConverter typeConverter;
   // Create a builder without an insertion point. We will insert using the
   // symbol table to guarantee unique names.
   OpBuilder globalBuilder(module.getContext());
   SymbolTable symbolTable(module);
if (it != globals.end())
       return;
 
     // Create a pretty name.
     SmallString<64> buf;
     llvm::raw_svector_ostream os(buf);
     interleave(type.getShape(), os, "x");
     os << "x" << type.getElementType();
 
-    auto global = globalBuilder.create<GlobalMemrefOp>(
+    auto global = globalBuilder.create<memref::GlobalOp>(
         op.getLoc(), (Twine("__constant_") + os.str()).str(),
         /*sym_visibility=*/globalBuilder.getStringAttr("private"),
         /*type=*/typeConverter.convertType(type),
         /*initial_value=*/op.getValue().cast<ElementsAttr>(),
         /*constant=*/true);
     symbolTable.insert(global);
     // The symbol table inserts at the end of the module, but globals are a bit
     // nicer if they are at the beginning.
public:
   LogicalResult
   matchAndRewrite(ConstantOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
     auto type = op.getType().dyn_cast<RankedTensorType>();
     if (!type)
       return failure();
 
     auto globalMemref = globals.getGlobalFor(op.value());
-    rewriter.replaceOpWithNewOp<GetGlobalMemrefOp>(op, globalMemref.type(),
+    rewriter.replaceOpWithNewOp<memref::GetGlobalOp>(op, globalMemref.type(),
                                                      globalMemref.getName());
     return success();
   }
   GlobalCreator &globals;
 };
 } // namespace
 
 namespace {
 struct TensorConstantBufferizePass
     : public TensorConstantBufferizeBase<TensorConstantBufferizePass> {
   void runOnOperation() override {
     auto module = getOperation();
     GlobalCreator globals(module);
 
     auto *context = &getContext();
     BufferizeTypeConverter typeConverter;
     OwningRewritePatternList patterns;
     ConversionTarget target(*context);
 
-    target.addLegalDialect<StandardOpsDialect>();
+    target.addLegalDialect<memref::MemRefDialect>();
     patterns.insert<BufferizeTensorConstantOp>(globals, typeConverter, context);
     target.addDynamicallyLegalOp<ConstantOp>(
         [&](ConstantOp op) { return typeConverter.isLegal(op.getType()); });
     if (failed(applyPartialConversion(module, target, std::move(patterns))))
       signalPassFailure();
   }
 };
 } // namespace
 
 std::unique_ptr<Pass> mlir::createTensorConstantBufferizePass() {
   return std::make_unique<TensorConstantBufferizePass>();
 }

mlir/lib/Transforms/Bufferize.cpp

 //===- Bufferize.cpp - Bufferization utilities ----------------------------===//
 //
 // 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/Transforms/Bufferize.h"
 #include "PassDetail.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/IR/Operation.h"
 #include "mlir/Transforms/Passes.h"
 
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
 // BufferizeTypeConverter
 //===----------------------------------------------------------------------===//
 
 static Value materializeTensorLoad(OpBuilder &builder, TensorType type,
                                    ValueRange inputs, Location loc) {
   assert(inputs.size() == 1);
   assert(inputs[0].getType().isa<BaseMemRefType>());
-  return builder.create<TensorLoadOp>(loc, type, inputs[0]);
+  return builder.create<memref::TensorLoadOp>(loc, type, inputs[0]);
 }
 
 /// Registers conversions into BufferizeTypeConverter
 BufferizeTypeConverter::BufferizeTypeConverter() {
   // Keep all types unchanged.
   addConversion([](Type type) { return type; });
   // Convert RankedTensorType to MemRefType.
   addConversion([](RankedTensorType type) -> Type {
     return MemRefType::get(type.getShape(), type.getElementType());
   });
   // Convert UnrankedTensorType to UnrankedMemRefType.
   addConversion([](UnrankedTensorType type) -> Type {
     return UnrankedMemRefType::get(type.getElementType(), 0);
   });
   addArgumentMaterialization(materializeTensorLoad);
   addSourceMaterialization(materializeTensorLoad);
   addTargetMaterialization([](OpBuilder &builder, BaseMemRefType type,
                               ValueRange inputs, Location loc) -> Value {
     assert(inputs.size() == 1);
     assert(inputs[0].getType().isa<TensorType>());
-    return builder.create<TensorToMemrefOp>(loc, type, inputs[0]);
+    return builder.create<memref::BufferCastOp>(loc, type, inputs[0]);
   });
 }
 
 void mlir::populateBufferizeMaterializationLegality(ConversionTarget &target) {
-  target.addLegalOp<TensorLoadOp, TensorToMemrefOp>();
+  target.addLegalOp<memref::TensorLoadOp, memref::BufferCastOp>();
 }
 
 namespace {
 // In a finalizing bufferize conversion, we know that all tensors have been
 // converted to memrefs, thus, this op becomes an identity.
-class BufferizeTensorLoadOp : public OpConversionPattern<TensorLoadOp> {
+class BufferizeTensorLoadOp : public OpConversionPattern<memref::TensorLoadOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
   LogicalResult
-  matchAndRewrite(TensorLoadOp op, ArrayRef<Value> operands,
+  matchAndRewrite(memref::TensorLoadOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
-    TensorLoadOp::Adaptor adaptor(operands);
+    memref::TensorLoadOp::Adaptor adaptor(operands);
     rewriter.replaceOp(op, adaptor.memref());
     return success();
   }
 };
 } // namespace
 
 namespace {
 // In a finalizing bufferize conversion, we know that all tensors have been
 // converted to memrefs, thus, this op becomes an identity.
-class BufferizeTensorToMemrefOp : public OpConversionPattern<TensorToMemrefOp> {
+class BufferizeCastOp : public OpConversionPattern<memref::BufferCastOp> {
 public:
   using OpConversionPattern::OpConversionPattern;
   LogicalResult
-  matchAndRewrite(TensorToMemrefOp op, ArrayRef<Value> operands,
+  matchAndRewrite(memref::BufferCastOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
-    TensorToMemrefOp::Adaptor adaptor(operands);
+    memref::BufferCastOp::Adaptor adaptor(operands);
     rewriter.replaceOp(op, adaptor.tensor());
     return success();
   }
 };
 } // namespace
 
 void mlir::populateEliminateBufferizeMaterializationsPatterns(
     MLIRContext *context, BufferizeTypeConverter &typeConverter,
     OwningRewritePatternList &patterns) {
-  patterns.insert<BufferizeTensorLoadOp, BufferizeTensorToMemrefOp>(
-      typeConverter, context);
+  patterns.insert<BufferizeTensorLoadOp, BufferizeCastOp>(typeConverter,
+                                                          context);
 }
 
 namespace {
 struct FinalizingBufferizePass
     : public FinalizingBufferizeBase<FinalizingBufferizePass> {
   using FinalizingBufferizeBase<
       FinalizingBufferizePass>::FinalizingBufferizeBase;
 

mlir/lib/Transforms/Canonicalizer.cpp

 //===- Canonicalizer.cpp - Canonicalize MLIR operations -------------------===//
 //
 // 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 converts operations into their canonical forms by
 // folding constants, applying operation identity transformations etc.
 //
 //===----------------------------------------------------------------------===//
 
 #include "PassDetail.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "mlir/Transforms/Passes.h"
 
 using namespace mlir;
 
 namespace {
 /// Canonicalize operations in nested regions.
 struct Canonicalizer : public CanonicalizerBase<Canonicalizer> {
   /// Initialize the canonicalizer by building the set of patterns used during
   /// execution.
   LogicalResult initialize(MLIRContext *context) override {
     OwningRewritePatternList owningPatterns;
     for (auto *op : context->getRegisteredOperations())
       op->getCanonicalizationPatterns(owningPatterns, context);
     patterns = std::move(owningPatterns);
     return success();
   }
   void runOnOperation() override {
     (void)applyPatternsAndFoldGreedily(getOperation()->getRegions(), patterns);
   }
 
   FrozenRewritePatternList patterns;
 };
 } // end anonymous namespace
 
 /// Create a Canonicalizer pass.
 std::unique_ptr<Pass> mlir::createCanonicalizerPass() {
   return std::make_unique<Canonicalizer>();
 }

mlir/lib/Transforms/Utils/Utils.cpp

 //===----------------------------------------------------------------------===//
 
 #include "mlir/Transforms/Utils.h"
 
 #include "mlir/Analysis/AffineAnalysis.h"
 #include "mlir/Analysis/AffineStructures.h"
 #include "mlir/Analysis/Utils.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/IR/Builders.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/Dominance.h"
 #include "mlir/Support/MathExtras.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/TypeSwitch.h"
 using namespace mlir;
 
if (domInstFilter && !domInfo->dominates(domInstFilter, op))
       continue;
 
     // Skip this use if it's not post-dominated by postDomInstFilter.
     if (postDomInstFilter && !postDomInfo->postDominates(postDomInstFilter, op))
       continue;
 
     // Skip dealloc's - no replacement is necessary, and a memref replacement
     // at other uses doesn't hurt these dealloc's.
-    if (isa<DeallocOp>(op) && !replaceInDeallocOp)
+    if (isa<memref::DeallocOp>(op) && !replaceInDeallocOp)
       continue;
 
     // Check if the memref was used in a non-dereferencing context. It is fine
     // for the memref to be used in a non-dereferencing way outside of the
     // region where this replacement is happening.
     if (!isa<AffineMapAccessInterface>(*op)) {
       if (!allowNonDereferencingOps)
         return failure();
for (unsigned i = 0, e = newOperands.size(); i < e; i++) {
     }
   }
   for (unsigned idx = 0, e = newOperands.size(); idx < e; idx++) {
     opInst->setOperand(idx, newOperands[idx]);
   }
 }
 
 // TODO: Currently works for static memrefs with a single layout map.
-LogicalResult mlir::normalizeMemRef(AllocOp allocOp) {
-  MemRefType memrefType = allocOp.getType();
-  OpBuilder b(allocOp);
+LogicalResult mlir::normalizeMemRef(memref::AllocOp *allocOp) {
+  MemRefType memrefType = allocOp->getType();
+  OpBuilder b(*allocOp);
 
   // Fetch a new memref type after normalizing the old memref to have an
   // identity map layout.
   MemRefType newMemRefType =
-      normalizeMemRefType(memrefType, b, allocOp.symbolOperands().size());
+      normalizeMemRefType(memrefType, b, allocOp->symbolOperands().size());
   if (newMemRefType == memrefType)
     // Either memrefType already had an identity map or the map couldn't be
     // transformed to an identity map.
     return failure();
 
-  Value oldMemRef = allocOp.getResult();
+  Value oldMemRef = allocOp->getResult();
 
-  SmallVector<Value, 4> symbolOperands(allocOp.symbolOperands());
-  AllocOp newAlloc = b.create<AllocOp>(allocOp.getLoc(), newMemRefType,
-                                       allocOp.alignmentAttr());
+  SmallVector<Value, 4> symbolOperands(allocOp->symbolOperands());
+  memref::AllocOp newAlloc = b.create<memref::AllocOp>(
+      allocOp->getLoc(), newMemRefType, allocOp->alignmentAttr());
   AffineMap layoutMap = memrefType.getAffineMaps().front();
   // Replace all uses of the old memref.
   if (failed(replaceAllMemRefUsesWith(oldMemRef, /*newMemRef=*/newAlloc,
                                       /*extraIndices=*/{},
                                       /*indexRemap=*/layoutMap,
                                       /*extraOperands=*/{},
                                       /*symbolOperands=*/symbolOperands,
                                       /*domInstFilter=*/nullptr,
                                       /*postDomInstFilter=*/nullptr,
                                       /*allowDereferencingOps=*/true))) {
     // If it failed (due to escapes for example), bail out.
     newAlloc.erase();
     return failure();
   }
   // Replace any uses of the original alloc op and erase it. All remaining uses
   // have to be dealloc's; RAMUW above would've failed otherwise.
-  assert(llvm::all_of(oldMemRef.getUsers(),
-                      [](Operation *op) { return isa<DeallocOp>(op); }));
+  assert(llvm::all_of(oldMemRef.getUsers(), [](Operation *op) {
+    return isa<memref::DeallocOp>(op);
+  }));
   oldMemRef.replaceAllUsesWith(newAlloc);
-  allocOp.erase();
+  allocOp->erase();
   return success();
 }
 
 MemRefType mlir::normalizeMemRefType(MemRefType memrefType, OpBuilder b,
                                      unsigned numSymbolicOperands) {
   unsigned rank = memrefType.getRank();
   if (rank == 0)
     return memrefType;