[mlir][Linalg] Update the semantics, verifier and test for Linalg with tensors.

[nicolasvasilache]

Summary

This diff fixes issues with the semantics of linalg.generic on tensors that appeared when converting directly from HLO to linalg.generic.
The changes are self-contained within MLIR and can be captured and tested independently of XLA.

The linalg.generic and indexed_generic are updated to:

To allow progressive lowering from the value world (a.k.a tensor values) to
the buffer world (a.k.a memref values), a linalg.generic op accepts
mixing input and output ranked tensor values with input and output memrefs.

%1 = linalg.generic #trait_attribute %A, %B {other-attributes} :
  tensor<?x?xf32>,
  memref<?x?xf32, stride_specification>
  -> (tensor<?x?xf32>)

In this case, the number of outputs (args_out) must match the sum of (1) the
number of output buffer operands and (2) the number of tensor return values.
The semantics is that the linalg.indexed_generic op produces (i.e.
allocates and fills) its return values.

Tensor values must be legalized by a buffer allocation pass before most
transformations can be applied. Such legalization moves tensor return values
into output buffer operands and updates the region argument accordingly.

Transformations that create control-flow around linalg.indexed_generic
operations are not expected to mix with tensors because SSA values do not
escape naturally. Still, transformations and rewrites that take advantage of
tensor SSA values are expected to be useful and will be added in the near
future.

[nicolasvasilache]

clang-format.

[merge_guards_bot]

Unit tests: pass. 61744 tests passed, 0 failed and 780 were skipped.

clang-tidy: fail. Please fix clang-tidy findings.

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[merge_guards_bot]

Unit tests: pass. 61744 tests passed, 0 failed and 780 were skipped.

clang-tidy: fail. Please fix clang-tidy findings.

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[herhut]

I browsed over the code and left some minor comments. The idea and code in verifiers LGTM. I find it much harder to read now with all those similar sounding accessors but that is inevitable I assume if you want to support mixed forms.

It seems that all transformation and code generation still assumes all buffer representations. That is fine (and also makes sense) but I'd prefer to have something emit an error or assert if this is not the case. One way less to shoot myself in the foot.

mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td

693–695

So this example has a single output assuming that there is an order constraint on the arguments (inputs first, then outputs). Could the describe this here?

mlir/include/mlir/Dialect/Linalg/IR/LinalgTraits.h

186

The number of results alone needs to the smaller than nOutputs(). Maybe assert that?

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

189

This iterators over all operands, including output buffers.

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

90–92

This code assumes an all buffer representation? Maybe add an assert?

228

This code assumes an all buffers representation?

[nicolasvasilache]

mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td

693–695

Updated the example, had forgotten this, thanks!

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

90–92

Sprinkled a bunch in the relevant places, thanks!

228

sprinkled a bunch of asserts in the relevant places, thanks!

[nicolasvasilache]

Address review comments.

[merge_guards_bot]

Unit tests: pass. 61794 tests passed, 0 failed and 781 were skipped.

clang-tidy: unknown.

clang-format: pass.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[ftynse]

mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td

60

Apologies for being obnoxious, but the review would be significantly faster if NFC code motion changes were separate from functional changes (in the former case I can quickly eyeball it to confirm it's indeed NFC, whereas in the late I need to understand deeply what the code does; without separation, the understanding time also extends on the NFC parts)

132

Why remove the documentation?

573–574

Unclear whether "allocates and fills" extends to both memrefs and tensors, or just tensors since only tensors are "return values".

578

Nit: argument -> arguments

580–582

I cannot understand what "does not mix with tensors" mean in this context. Consider rephrasing or providing an example.

698–702

I wonder if you could write this text only once in a tablegen variable for both linalg.generic and linalg.indexed_generic, and then just concatenate that variable with the rest of the op-specific documentation. This would avoid the duplication and everything that comes with it, but may make it harder for a casual reader of the op definition to follow. Since you are the only person actively supporting this and doing the work twice, your call.

mlir/include/mlir/Dialect/Linalg/IR/LinalgTraits.h

123–124

Nit: values->buffers

136

Typo "depending regardless"

138

Nit: remove this //, I was confused by it to read the block comment above as describing the function below, which is contradictory.

188

Spurious semicolon

mlir/lib/Dialect/Linalg/Analysis/DependenceAnalysis.cpp

142–146

This dependence analysis now only works if the output is a buffer, and cannot be queried if it is a tensor. Consider documenting (and maybe extending later to also work for tensors).

148

Do you need any modifications to support tensor inputs here? From a quick look, it should just work, but better check...

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

186

Ultra-nit: functions are not necessarily fun, please use "function" in user-visible messages

188–201

Why +1 here? It's inconsistent with block arguments above.

220

The use of 0-based index here is inconsistent with the 1-based index below. Please adopt a convention, document it in the dialect documentation and use everywhere.

225–239

This looks common with then non-indexed version. Can this be factored out into a helper function?

243

Nit: "inputs and buffer operands" sounds like a false dichotomy.

728

Ultra-nit: whitespace before (

[nicolasvasilache]

Address review comments.

mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td

60

Apologies for this reordering.
I was contemplating extracting and landing an NFC but please note that the code that moved is clearly marked by phabricator with the yellow vertical bar.
When you hover over it it says Moved from line xxx.

If you think that is still not enough I can rebase the NFC part but I personally have found phabricator to be good for this.

132

Because the proper place for this doc is inside the op interface and it is already there, almost word for word.

698–702

Will consider for a followup, thanks!

mlir/lib/Dialect/Linalg/Analysis/DependenceAnalysis.cpp

142–146

Added an assertion, there is no short term plan to support such transformations and analyses on tensors.
The answer is buffer allocate then it's in the buffer world and these apply.
Mixed mode may appear in the future but would require rethinking things.

148

added assertions above.

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

188–201

made it consistent, error messages for function arguments start at 1: e.g. 1st argument etc.

220

Made it consistent, thanks for spotting!

[merge_guards_bot]

Unit tests: pass. 61794 tests passed, 0 failed and 781 were skipped.

clang-tidy: unknown.

clang-format: pass.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[ftynse]

Something went wrong with formatting in the commit description (mixed double space prefix and backticks?). Please reformat and feel free to land.

[nicolasvasilache]

Rebasing.

[merge_guards_bot]

Unit tests: pass. 61863 tests passed, 0 failed and 781 were skipped.

clang-tidy: unknown.

clang-format: pass.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[asaadaldien]

mlir/include/mlir/Dialect/Linalg/IR/LinalgTraits.h

204

Do we need to assert(i - getNumInputsAndOutputBuffers() < getOutputTensorTypes().size()) ?

mlir/include/mlir/Dialect/Linalg/IR/LinalgStructuredOps.td

class NOutputs<int args_out> :
   NativeOpTrait<"linalg::NOutputs<" # !cast<string>(args_out) # ">::Impl"> {}
 
 def StructuredOpTraits : NativeOpTrait<"linalg::StructuredOpTraits">;
 
 // The linalg 'LinalgStructuredInterface' provides access to the 'LinalgOp'
 // interface.
 def LinalgStructuredInterface : OpInterface<"LinalgOp"> {
   let methods = [
-    InterfaceMethod<
-      "Query the number of inputs from the current operation.",
-      "unsigned", "getNumInputs"
-    >,
-    InterfaceMethod<
-      "Query the number of outputs from the current operation.",
-      "unsigned", "getNumOutputs"
-    >,
-    InterfaceMethod<
-      "Query the number of inputs and outputs from the current operation.",
-      "unsigned", "getNumInputsAndOutputs"
-    >,
-    InterfaceMethod<
-      "Query the input operands from the current operation.",
-      "Operation::operand_range", "getInputs"
-    >,
-    InterfaceMethod<
-      "Query the output operands from the current operation.",
-      "Operation::operand_range", "getOutputs"
-    >,
-    InterfaceMethod<
-      "Query the input and output operands from the current operation.",
-      "Operation::operand_range", "getInputsAndOutputs"
-    >,
-    InterfaceMethod<
-      "Query the iterator types attribute within the current operation.",
-      "ArrayAttr", "iterator_types"
-    >,
-    InterfaceMethod<
-      "Query the indexing maps attribute within the current operation.",
-      "ArrayAttr", "indexing_maps"
-    >,
+    //========================================================================//
+    // Loop types handling.
+    //========================================================================//
     InterfaceMethod<
       "Query the number of parallel loops within the current operation.",
       "unsigned", "getNumParallelLoops"
     >,
     InterfaceMethod<
       "Query the number of reduction loops within the current operation.",
       "unsigned", "getNumReductionLoops"
     >,
     InterfaceMethod<
       "Query the number of window loops within the current operation.",
       "unsigned", "getNumWindowLoops"
     >,
     InterfaceMethod<
       "Query the number of loops within the current operation.",
       "unsigned", "getNumLoops">,
+
+    //========================================================================//
+    // Input arguments handling.
+    //========================================================================//
+    InterfaceMethod<
+      "Query the number of inputs from the current operation.",
+      "unsigned", "getNumInputs"
+    >,
     InterfaceMethod<"Query the input view at the given index.",
       "Value ", "getInput", (ins "unsigned":$i)
     >,
-    InterfaceMethod<"Query the output view at the given index.",
-      "Value ", "getOutput", (ins "unsigned":$i)
-    >,
     InterfaceMethod<[{
         Return the index of the given input value `v`, or `None` if the value is
         not an input.
       }],
       "llvm::Optional<unsigned>", "getIndexOfInput", (ins "Value ":$v)
     >,
-    InterfaceMethod<[{
-        Query the index of the given view value, or `None` if the value is not
-        a view.
-      }],
-      "llvm::Optional<unsigned>", "getIndexOfOutput", (ins "Value ":$view)
+    InterfaceMethod<
+      "Query the input operands from the current operation.",
+      "Operation::operand_range", "getInputs"
     >,
     InterfaceMethod<[{
         Query the type of the input shape at the given index.
       }], "ShapedType", "getInputShapedType", (ins "unsigned":$i)>,
     InterfaceMethod<[{
-        Query the type of the output view at the given index.
-      }], "ShapedType", "getOutputShapedType", (ins "unsigned":$i)>,
-    InterfaceMethod<[{
-        Query whether the op has only MemRef input and outputs.
-      }], "bool", "hasBufferSemantics">,
-    InterfaceMethod<[{
         Query the subset of input operands that are of ranked tensor type.
       }], "SmallVector<RankedTensorType, 4>", "getInputTensorTypes">,
+
+
+    //========================================================================//
+    // Output arguments handling.
+    //========================================================================//
+    InterfaceMethod<
+      "Query the number of outputs from the current operation.",
+      "unsigned", "getNumOutputs"
+    >,
+    InterfaceMethod<"Query the output buffer at the given index.",
+      "Value ", "getOutputBuffer", (ins "unsigned":$i)
+    >,
+    InterfaceMethod<[{
+        Query the index of the given buffer value, or `None` if the value is not
+        part of the output buffers.
+      }],
+      "llvm::Optional<unsigned>", "getIndexOfOutputBuffer", (ins "Value ":$view)
+    >,
+    InterfaceMethod<[{
+        Query the type of the output buffer at the given index.
+      }], "MemRefType", "getOutputBufferType", (ins "unsigned":$i)>,
     InterfaceMethod<[{
-        Query the subset of output operands that are of ranked tensor type.
+        Query the results that are of ranked tensor type.
       }], "SmallVector<RankedTensorType, 4>", "getOutputTensorTypes">,
+    InterfaceMethod<
+      "Query the output buffers (operands) from the current operation.",
+      "Operation::operand_range", "getOutputBuffers"
+    >,
+
+    //========================================================================//
+    // Input and Output arguments handling.
+    //========================================================================//
+    InterfaceMethod<
+      "Return the number of inputs and outputs, irrespective of their buffer "
+      "or tensor type.",
+      "unsigned", "getNumInputsAndOutputs"
+    >,
+    InterfaceMethod<
+      "Return the number of inputs, irrespective of their buffer or tensor "
+      "type, and output buffers",
+      "unsigned", "getNumInputsAndOutputBuffers"
+    >,
+    InterfaceMethod<
+      "Return the range over inputs (irrespective of type) and output buffers.",
+      "Operation::operand_range", "getInputsAndOutputBuffers"
+    >,
 
+    //========================================================================//
+    // Other interface methods.
+    //========================================================================//
+    InterfaceMethod<
+      "Query the iterator types attribute within the current operation.",
+      "ArrayAttr", "iterator_types"
+    >,
+    InterfaceMethod<
+      "Query the indexing maps attribute within the current operation.",
+      "ArrayAttr", "indexing_maps"
+    >,
+    InterfaceMethod<[{
+        Query whether the op has only MemRef input and outputs.
+      }], "bool", "hasBufferSemantics">,
+
+    //========================================================================//
+    // Other static interface methods.
+    //========================================================================//
     StaticInterfaceMethod<[{
         Create an operation of the current type with the given location,
         operands, and attributes.
       }],
       "Operation *", "create",
       (ins "OpBuilder &":$builder, "Location":$loc,
            "ValueRange":$operands,
            "ArrayRef<NamedAttribute>":$attributes), [{
         return builder.create<ConcreteOp>(loc, ArrayRef<Type>{}, operands,
                                           attributes);
       }]
     >,
-
-    /// Clone an operation with the given location and operands. This is used to
-    /// abstract away the optional underlying region creation.
     InterfaceMethod<[{
         Clone the current operation with the given location and operands. This
         is used to abstract away the optional underlying region creation.
       }],
       "Operation *", "clone",
       (ins "OpBuilder &":$b, "Location":$loc, "ValueRange":$operands), [{
         BlockAndValueMapping map;
         unsigned numRegions = op.getOperation()->getNumRegions();
let description = [{
     }
     ```
 
     To allow progressive lowering from the value world (a.k.a tensor values) to
     the buffer world (a.k.a memref values), a `linalg.generic` op accepts
     mixing input and output ranked tensor values with input and output memrefs.
 
     ```mlir
-      %1 = linalg.generic #trait_attribute %A, %B, %C {other-attributes} :
+      %C = linalg.generic #trait_attribute %A, %B {other-attributes} :
         tensor<?x?xf32>,
-        memref<?x?xf32, stride_specification>,
-        tensor<?x?xf32>
+        memref<?x?xf32, stride_specification>
         -> (tensor<?x?xf32>)
     ```
 
-    In this case, the number of return values must match the number of output
-    tensor arguments. The semantics is that the `linalg.generic` op
-    produces (i.e. allocates and fills) its return values.
+    In this case, the number of outputs (args_out) must match the sum of (1) the
+    number of output buffer operands and (2) the number of tensor return values.
+    The semantics is that the `linalg.indexed_generic` op produces (i.e.
+    allocates and fills) its tensor return values.
+
     Tensor values must be legalized by a buffer allocation pass before most
-    transformations can be applied. In particular, transformations that create
-    control flow around linalg.generic operations are not expected to mix with
-    tensors because SSA values do not escape naturally. Still, transformations
-    and rewrites that take advantage of tensor SSA values are expected to be
-    useful and will be added in the near future.
+    transformations can be applied. Such legalization moves tensor return values
+    into output buffer operands and updates the region arguments accordingly.
+
+    Transformations that create control-flow around linalg.indexed_generic
+    operations are not expected to work with tensors because SSA values do not
+    escape naturally. Still, transformations and rewrites that take advantage of
+    tensor SSA values are expected to be useful and will be added in the near
+    future.
   }];
   let verifier = [{ return ::verify(*this); }];
 }
 
 def IndexedGenericOp : GenericOpBase<"indexed_generic"> {
   let description = [{
     Indexed Generic Linalg op form where the key properties of the computation
     are specified as attributes. In pretty form, a linalg.indexed_generic op is
let description = [{
     ```
 
     To allow progressive lowering from the value world (a.k.a tensor values) to
     the buffer world (a.k.a memref values), a `linalg.indexed_generic` op
     accepts mixing input and output ranked tensor values with input and output
     memrefs.
 
     ```mlir
-      %1 = linalg.indexed_generic #trait_attribute %A, %B, %C {other-attributes}
+      %C = linalg.indexed_generic #trait_attribute %A, %B {other-attributes}
       : tensor<?x?xf32>,
-        memref<?x?xf32, stride_specification>,
-        tensor<?x?xf32>
+        memref<?x?xf32, stride_specification>
         -> (tensor<?x?xf32>)
     ```
 
-    In this case, the number of return values must match the number of output
-    tensor arguments. The semantics is that the `linalg.indexed_generic` op
-    produces (i.e. allocates and fills) its return values.
+    In this case, the number of outputs (args_out) must match the sum of (1) the
+    number of output buffer operands and (2) the number of tensor return values.
+    The semantics is that the `linalg.indexed_generic` op produces (i.e.
+    allocates and fills) its return values.
+
     Tensor values must be legalized by a buffer allocation pass before most
-    transformations can be applied. In particular, transformations that create
-    control flow around linalg.generic operations are not expected to mix with
-    tensors because SSA values do not escape naturally. Still, transformations
-    and rewrites that take advantage of tensor SSA values are expected to be
-    useful and will be added in the near future.
+    transformations can be applied. Such legalization moves tensor return values
+    into output buffer operands and updates the region argument accordingly.
+
+    Transformations that create control-flow around linalg.indexed_generic
+    operations are not expected to work with tensors because SSA values do not
+    escape naturally. Still, transformations and rewrites that take advantage of
+    tensor SSA values are expected to be useful and will be added in the near
+    future.
   }];
   let verifier = [{ return ::verify(*this); }];
 }
 
 #endif // LINALG_STRUCTURED_OPS

mlir/include/mlir/Dialect/Linalg/IR/LinalgTraits.h

 /// `getNumOutputs`. Use as a trait as follows:
 ///
 ///   class DotOp : public Op<DotOp, OpTrait::StructuredOpTraits> {
 ///
 template <typename ConcreteType>
 class StructuredOpTraits
     : public OpTrait::TraitBase<ConcreteType, StructuredOpTraits> {
 private:
-  /// Return the number of inputs. For internal use only.
+  /// Return the number of inputs, irrespective of their buffer or tensor type.
+  /// For internal use only.
   unsigned nInputs() {
     return cast<ConcreteType>(this->getOperation()).getNumInputs();
   }
-  /// Return the number of outputs. For internal use only.
+  /// Return the number of outputs, irrespective of their buffer or tensor type.
+  /// For internal use only.
   unsigned nOutputs() {
     return cast<ConcreteType>(this->getOperation()).getNumOutputs();
   }
 
 public:
+  //==========================================================================//
+  // Loop types handling.
+  //==========================================================================//
+  unsigned getNumParallelLoops() {
+    return getNumIterators(
+        getParallelIteratorTypeName(),
+        cast<ConcreteType>(this->getOperation()).iterator_types());
+  }
+  unsigned getNumReductionLoops() {
+    return getNumIterators(
+        getReductionIteratorTypeName(),
+        cast<ConcreteType>(this->getOperation()).iterator_types());
+  }
+  unsigned getNumWindowLoops() {
+    return getNumIterators(
+        getWindowIteratorTypeName(),
+        cast<ConcreteType>(this->getOperation()).iterator_types());
+  }
+  unsigned getNumLoops() {
+    return getNumIterators(
+        cast<ConcreteType>(this->getOperation()).iterator_types());
+  }
+
+  //==========================================================================//
+  // Input arguments handling.
+  //==========================================================================//
+  // The `i^th` input argument is always the `i^th` operand regardless of
+  // whether we have tensors or buffers.
+  //
   /// Return the `i`-th input value.
   Value getInput(unsigned i) {
     assert(i < nInputs());
     return this->getOperation()->getOperand(i);
   }
   /// Return the index of `value` in the list of inputs if found, llvm::None
   /// otherwise.
   Optional<unsigned> getIndexOfInput(Value value) {
     auto it = llvm::find(getInputs(), value);
     if (it != getInputs().end())
       return it - getInputs().begin();
     return llvm::None;
   }
   /// Return the `i`-th input buffer type.
   ShapedType getInputShapedType(unsigned i) {
     return getInput(i).getType().template cast<ShapedType>();
   }
   /// Return the range over inputs.
   Operation::operand_range getInputs() {
     auto range = this->getOperation()->getOperands();
     return {range.begin(), range.begin() + nInputs()};
   }
-  /// Return the `i`-th output.
-  Value getOutput(unsigned i) {
-    return this->getOperation()->getOperand(nInputs() + i);
-  }
-  /// Return the index of `value` in the list of output values if found,
-  /// llvm::None otherwise.
-  Optional<unsigned> getIndexOfOutput(Value value) {
-    auto it = llvm::find(getOutputs(), value);
-    if (it != getOutputs().end())
-      return it - getOutputs().begin();
-    return llvm::None;
-  }
-  /// Return the `i`-th output buffer type.
-  ShapedType getOutputShapedType(unsigned i) {
-    return getOutput(i).getType().template cast<ShapedType>();
-  }
-  /// Query whether the op has only MemRef input and outputs.
-  bool hasBufferSemantics() {
-    return this->getOperation()->getNumResults() == 0 &&
-           llvm::all_of(getInputsAndOutputs(),
-                        [](Value v) { return v.getType().isa<MemRefType>(); });
-  }
   /// Query the subset of input operands that are of ranked tensor type.
   SmallVector<RankedTensorType, 4> getInputTensorTypes() {
     SmallVector<RankedTensorType, 4> res;
     for (Type type : getInputs().getTypes())
       if (auto t = type.template dyn_cast<RankedTensorType>())
         res.push_back(t);
     return res;
   }
-  /// Query the subset of output operands that are of ranked tensor type.
+
+  //==========================================================================//
+  // Output arguments handling.
+  //==========================================================================//
+  // The `i^th` output argument is an operand (resp. a return value) iff it is
+  // a value of buffer type (resp. a return value of tensor type).
+
+  /// Return the `i`-th output, asserts that this is a buffer operand and not
+  /// a tensor result.
+  Value getOutputBuffer(unsigned i) {
+    assert(i + this->getOperation()->getNumResults() < nOutputs() &&
+           "overflowing output buffer index");
+    return this->getOperation()->getOperand(nInputs() + i);
+  }
+  /// Return the index of `value` in the list of output buffers if found,
+  /// llvm::None otherwise.
+  Optional<unsigned> getIndexOfOutputBuffer(Value value) {
+    auto it = llvm::find(getOutputBuffers(), value);
+    if (it != getOutputBuffers().end())
+      return it - getOutputBuffers().begin();
+    return llvm::None;
+  }
+  /// Return the `i`-th output buffer type.
+  MemRefType getOutputBufferType(unsigned i) {
+    return getOutputBuffer(i).getType().template cast<MemRefType>();
+  }
+  /// Return the `i`-th output shaped type, irrespective of buffer of tensor
+  /// type.
+  ShapedType getOutputShapedType(unsigned i) {
+    return getShapedType(i + nInputs());
+  }
+  /// Query the subset of results that are of ranked tensor type.
   SmallVector<RankedTensorType, 4> getOutputTensorTypes() {
     SmallVector<RankedTensorType, 4> res;
-    for (Type type : getOutputs().getTypes())
-      if (auto t = type.template dyn_cast<RankedTensorType>())
-        res.push_back(t);
+    for (Type type : this->getOperation()->getResults().getTypes())
+      res.push_back(type.template cast<RankedTensorType>());
     return res;
   }
   /// Return the range over outputs.
-  Operation::operand_range getOutputs() {
+  Operation::operand_range getOutputBuffers() {
     auto range = this->getOperation()->getOperands();
     return {range.begin() + nInputs(),
-            range.begin() + getNumInputsAndOutputs()};
+            range.begin() + getNumInputsAndOutputBuffers()};
   }
-  /// Return the number of inputs and outputs.
+
+  //==========================================================================//
+  // Input and Output arguments handling.
+  //==========================================================================//
+  /// Return the number of inputs and outputs, irrespective of their buffer or
+  /// tensor type.
   unsigned getNumInputsAndOutputs() { return nInputs() + nOutputs(); }
-  /// Return the `i`-th buffer type.
-  ShapedType getShapedType(unsigned i) {
-    return (i < nInputs()) ? getInputShapedType(i)
-                           : getOutputShapedType(i - nInputs());
+  /// Return the number of inputs, irrespective of their buffer or tensor type,
+  /// and output buffers.
+  unsigned getNumInputsAndOutputBuffers() {
+    assert(this->getOperation()->getNumResults() <= nOutputs());
+    return nInputs() + nOutputs() - this->getOperation()->getNumResults();
   }
-  /// Return the range over inputs and outputs.
-  Operation::operand_range getInputsAndOutputs() {
+  /// Return the range over inputs (irrespective of type) and output buffers.
+  Operation::operand_range getInputsAndOutputBuffers() {
     auto range = this->getOperation()->getOperands();
-    return {range.begin(), range.begin() + getNumInputsAndOutputs()};
+    return {range.begin(), range.begin() + getNumInputsAndOutputBuffers()};
   }
-  unsigned getNumParallelLoops() {
-    return getNumIterators(
-        getParallelIteratorTypeName(),
-        cast<ConcreteType>(this->getOperation()).iterator_types());
-  }
-  unsigned getNumReductionLoops() {
-    return getNumIterators(
-        getReductionIteratorTypeName(),
-        cast<ConcreteType>(this->getOperation()).iterator_types());
-  }
-  unsigned getNumWindowLoops() {
-    return getNumIterators(
-        getWindowIteratorTypeName(),
-        cast<ConcreteType>(this->getOperation()).iterator_types());
+  /// Return the `i`-th shaped type, there are 3 cases:
+  ///   1. if `i < nInputs()` then return `getInputShapedType(i)`; otherwise
+  ///   2. if `i < getNumInputsAndOutputBuffers()` then return the
+  ///      `getOutputBufferType(i - nInputs())`; otherwise
+  ///   3. return the `i - getNumInputsAndOutputBuffers()` result type.
+  ShapedType getShapedType(unsigned i) {
+    if (i < nInputs())
+      return getInputShapedType(i);
+    if (i < getNumInputsAndOutputBuffers())
+      return getOutputBufferType(i - nInputs()).template cast<ShapedType>();
+    return getOutputTensorTypes()[i - getNumInputsAndOutputBuffers()]
+        .template cast<ShapedType>();
   }
-  unsigned getNumLoops() {
-    return getNumIterators(
-        cast<ConcreteType>(this->getOperation()).iterator_types());
+
+  //==========================================================================//
+  // Other interface methods.
+  //==========================================================================//
+  /// Query whether the op has only buffer inputs and no returns.
+  bool hasBufferSemantics() {
+    return this->getOperation()->getNumResults() == 0 &&
+           llvm::all_of(getInputs(),
+                        [](Value v) { return v.getType().isa<MemRefType>(); });
   }
+
+  //==========================================================================//
+  // Other static interface methods.
+  //==========================================================================//
   static LogicalResult verifyTrait(Operation *op) {
-    auto nOperands = cast<ConcreteType>(op).getNumInputsAndOutputs();
+    auto nOperands = cast<ConcreteType>(op).getNumInputsAndOutputBuffers();
     if (failed(OpTrait::impl::verifyAtLeastNOperands(op, nOperands)))
       return failure();
     return success();
   }
 };
 
 } // namespace linalg
 } // namespace OpTrait
 } // namespace mlir
 
 #endif // MLIR_DIALECT_LINALG_LINALGTRAITS_H_

mlir/include/mlir/Dialect/Linalg/Utils/Utils.h

Optional<FusionInfo> fuseProducerOf(OpBuilder &b, LinalgOp consumer,
                                     OperationFolder *folder = nullptr);
 
 /// Returns the linearized list of all view dimensions in a linalgOp. Applying
 /// the inverse, concatenated loopToOperandRangeMaps to this list allows the
 /// derivation of loop ranges for any linalgOp.
 template <typename ConcreteOp>
 SmallVector<Value, 8> getViewSizes(ConcreteOp linalgOp) {
   SmallVector<Value, 8> res;
-  for (auto v : linalgOp.getInputsAndOutputs()) {
+  for (auto v : linalgOp.getInputsAndOutputBuffers()) {
     MemRefType t = v.getType().template cast<MemRefType>();
     for (unsigned i = 0; i < t.getRank(); ++i)
       res.push_back(edsc::intrinsics::dim(v, i));
   }
   return res;
 }
 
 /// Returns the values obtained by applying `map` to the list of values.

mlir/lib/Dialect/Linalg/Analysis/DependenceAnalysis.cpp

LinalgDependenceGraph::getDependencesInto(
     Operation *dst, LinalgDependenceGraph::DependenceType dt) const {
   auto iter = dependencesIntoGraphs[dt].find(dst);
   if (iter == dependencesIntoGraphs[dt].end())
     return llvm::make_range(nullptr, nullptr);
   return llvm::make_range(iter->second.begin(), iter->second.end());
 }
 
 void LinalgDependenceGraph::addDependencesBetween(LinalgOp src, LinalgOp dst) {
-  for (auto srcView : src.getOutputs()) { // W
+  assert(src.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  assert(dst.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  for (auto srcView : src.getOutputBuffers()) { // W
     // RAW graph
     for (auto dstView : dst.getInputs()) {   // R
       if (aliases.alias(srcView, dstView)) { // if alias, fill RAW
         addDependenceElem(DependenceType::RAW,
                           LinalgOpView{src.getOperation(), srcView},
                           LinalgOpView{dst.getOperation(), dstView});
       }
     }
     // WAW graph
-    for (auto dstView : dst.getOutputs()) {  // W
+    for (auto dstView : dst.getOutputBuffers()) { // W
       if (aliases.alias(srcView, dstView)) { // if alias, fill WAW
         addDependenceElem(DependenceType::WAW,
                           LinalgOpView{src.getOperation(), srcView},
                           LinalgOpView{dst.getOperation(), dstView});
       }
     }
   }
   for (auto srcView : src.getInputs()) { // R
     // RAR graph
     for (auto dstView : dst.getInputs()) {   // R
       if (aliases.alias(srcView, dstView)) { // if alias, fill RAR
         addDependenceElem(DependenceType::RAR,
                           LinalgOpView{src.getOperation(), srcView},
                           LinalgOpView{dst.getOperation(), dstView});
       }
     }
     // WAR graph
-    for (auto dstView : dst.getOutputs()) {  // W
+    for (auto dstView : dst.getOutputBuffers()) { // W
       if (aliases.alias(srcView, dstView)) { // if alias, fill WAR
         addDependenceElem(DependenceType::WAR,
                           LinalgOpView{src.getOperation(), srcView},
                           LinalgOpView{dst.getOperation(), dstView});
       }
     }
   }
 }

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

static ParseResult parseGenericOp(OpAsmParser &parser, OperationState &result) {
   return parser.resolveOperands(operandsInfo, operandTypes,
                                 parser.getCurrentLocation(), result.operands);
 }
 
 template <typename GenericOpType>
 LogicalResult verifyBlockArgs(GenericOpType op, Block &block);
 
 template <> LogicalResult verifyBlockArgs(GenericOp op, Block &block) {
-  auto nViews = op.getNumInputsAndOutputs();
-  auto nInputViews = op.getNumInputs();
-  if (block.getNumArguments() != nViews)
-    return op.emitOpError(
-        "expected number of block arguments to match number of views");
+  auto nOperands = op.getNumOperands();
+  if (block.getNumArguments() != nOperands)
+    return op.emitOpError("expected number of block arguments to match number "
+                          "of operands");
 
-  for (unsigned i = 0; i < nViews; ++i) {
+  // Note: the number and type of yield values are checked in the YieldOp.
+  auto nInputViews = op.getNumInputs();
+  for (unsigned i = 0; i < nOperands; ++i) {
     auto viewType = op.getShapedType(i);
     if (viewType.getElementType() != block.getArgument(i).getType())
       return op.emitOpError("expected block argument ")
-             << i << " of the same type as elemental type of "
+             << (i + 1) << " of the same type as elemental type of "
              << ((i < nInputViews) ? "input " : "output ")
-             << "view: " << viewType;
+             << "operand: " << viewType;
   }
   return success();
 }
 
 template <> LogicalResult verifyBlockArgs(IndexedGenericOp op, Block &block) {
   auto nInputViews = op.getNumInputs();
   auto nLoops = op.getNumLoops();
-  auto nViews = op.getNumInputsAndOutputs();
-  if (block.getNumArguments() != nViews + nLoops)
+  auto nOperands = op.getNumOperands();
+  if (block.getNumArguments() != nOperands + nLoops)
     return op.emitOpError(
-        "expected number of block arguments to match number of views + "
+        "expected number of block arguments to match number of operands + "
         "number of loops");
 
-  for (unsigned i = 0; i < nLoops; ++i) {
+  // Note: the number and type of yield values are checked in the YieldOp.
+  for (unsigned i = 0; i < nLoops; ++i)
     if (!block.getArgument(i).getType().isIndex())
       return op.emitOpError("expected block argument ")
-             << i << " to be of IndexType";
-  }
+             << (i + 1) << " to be an index";
 
-  for (unsigned i = 0; i < nViews; ++i) {
+  for (unsigned i = 0; i < nOperands; ++i) {
     unsigned memrefArgIndex = i + nLoops;
     auto viewType = op.getShapedType(i);
     if (viewType.getElementType() !=
         block.getArgument(memrefArgIndex).getType())
       return op.emitOpError("expected block argument ")
-             << memrefArgIndex << " of the same type as elemental type of "
+             << (memrefArgIndex + 1)
+             << " of the same type as elemental type of "
              << ((i < nInputViews) ? "input " : "output ")
-             << "view: " << viewType;
+             << "operand: " << viewType;
   }
   return success();
 }
 
 template <typename GenericOpType>
 LogicalResult verifyFuncArgs(GenericOpType op, FunctionType funType);
 
-template <> LogicalResult verifyFuncArgs(GenericOp op, FunctionType funType) {
-  auto nViews = op.getNumInputsAndOutputs();
-  auto nInputViews = op.getNumInputs();
-  if (funType.getNumInputs() != nViews)
-    return op.emitOpError("expected fun arguments to match number of views");
-  if (funType.getNumResults() != op.getNumOutputs())
-    return op.emitOpError(
-        "expected fun results to match number of output views");
+template <typename GenericOpType>
+LogicalResult verifyFuncArgsGeneric(GenericOpType op, FunctionType funType) {
+  auto res = verifyFuncArgs(op, funType);
+  if (failed(res))
+    return res;
 
-  for (auto en : llvm::enumerate(op.indexing_maps())) {
-    auto idx = en.index();
-    auto view = (idx < nInputViews) ? op.getInputShapedType(idx)
-                                    : op.getOutputShapedType(idx - nInputViews);
-    if (funType.getInput(idx) != view.getElementType())
-      return op.emitOpError("expected fun argument ")
-             << idx << " of the same type as elemental type "
-             << view.getElementType() << " of view " << idx;
-
-    if (idx >= nInputViews) {
-      auto resultIdx = idx - nInputViews;
-      if (funType.getResult(resultIdx) != view.getElementType())
-        return op.emitOpError("expected fun result ")
-               << resultIdx << " of the same type as elemental type "
-               << view.getElementType() << " of view " << idx;
-    }
+  auto nInputs = op.getNumInputs();
+  auto nOutputs = op.getNumOutputs();
+  // linalg.generic output element types are exactly the function results.
+  for (unsigned idx = 0; idx < nOutputs; ++idx) {
+    ShapedType shapedType = op.getShapedType(nInputs + idx);
+    if (funType.getResult(idx) != shapedType.getElementType())
+      return op.emitOpError("expected function result ")
+             << (idx + 1) << " of the same type as elemental type "
+             << shapedType.getElementType() << " of output " << (idx + 1);
+  return success();
+}
+
+template <> LogicalResult verifyFuncArgs(GenericOp op, FunctionType funType) {
+  auto nOperands = op.getNumOperands();
+  if (funType.getNumInputs() != nOperands)
+    return op.emitOpError(
+        "expected function arguments to match number of operands");
+  if (funType.getNumResults() != op.getNumOutputs())
+    return op.emitOpError("expected function results(")
+           << funType.getNumResults() << ") to match number of outputs("
+           << op.getNumOutputs() << ")";
+
+  // linalg.generic operands element types are exactly the first function
+  // arguments.
+  for (unsigned idx = 0; idx < nOperands; ++idx) {
+    ShapedType shapedType = op.getShapedType(idx);
+    if (funType.getInput(idx) != shapedType.getElementType())
+      return op.emitOpError("expected function argument ")
+             << (idx + 1) << " of the same type as elemental type "
+             << shapedType.getElementType() << " of operand " << (idx + 1);
   }
+
   return success();
 }
 
 template <>
 LogicalResult verifyFuncArgs(IndexedGenericOp op, FunctionType funType) {
   auto nLoops = op.getNumLoops();
-  auto nInputViews = op.getNumInputs();
   auto nOutputs = op.getNumOutputs();
-  auto nViews = op.getNumInputsAndOutputs();
-  if (funType.getNumInputs() != nViews + nLoops)
-    return op.emitOpError(
-        "expected fun arguments to match number of views + number of loops");
+  auto nOperands = op.getNumOperands();
+  if (funType.getNumInputs() != nOperands + nLoops)
+    return op.emitOpError("expected function arguments to match number of "
+                          "loops + number of operands");
   if (funType.getNumResults() != nOutputs)
     return op.emitOpError(
-        "expected fun results to match number of output views");
-  for (unsigned i = 0; i < nLoops; ++i) {
+        "expected function results to match number of outputs");
+  for (unsigned i = 0; i < nLoops; ++i)
     if (!funType.getInput(i).isIndex())
-      return op.emitOpError("expected fun argument ")
-             << i << " to be of IndexType";
-  }
-  for (auto en : llvm::enumerate(op.indexing_maps())) {
-    auto idx = en.index();
-    auto funIdx = nLoops + idx;
-    auto view = (idx < nInputViews) ? op.getInputShapedType(idx)
-                                    : op.getOutputShapedType(idx - nInputViews);
-    if (funType.getInput(funIdx) != view.getElementType())
-      return op.emitOpError("expected fun argument ")
-             << funIdx << " of the same type as elemental type "
-             << view.getElementType() << " of view " << idx;
-
-    if (idx >= nInputViews) {
-      auto resultIdx = idx - nInputViews;
-      if (funType.getResult(resultIdx) != view.getElementType())
-        return op.emitOpError("expected fun result ")
-               << resultIdx << " of the same type as elemental type "
-               << view.getElementType() << " of view " << idx;
-    }
+      return op.emitOpError("expected function argument ")
+             << (i + 1) << " to be an index";
+
+  // linalg.generic operands element types are exactly the first function
+  // arguments.
+  for (unsigned idx = 0; idx < nOperands; ++idx) {
+    ShapedType shapedType = op.getShapedType(idx);
+    if (funType.getInput(idx + nLoops) != shapedType.getElementType())
+      return op.emitOpError("expected function argument ")
+             << (idx + nLoops + 1) << " of the same type as elemental type "
+             << shapedType.getElementType() << " of input " << (idx + 1);
   }
+
   return success();
 }
 
 template <typename GenericOpType>
 LogicalResult verifyGenericOp(GenericOpType op) {
   auto nInputViews = op.getNumInputs();
   auto nLoops = op.getNumLoops();
-  auto nViews = op.getNumInputsAndOutputs();
-  if (nViews != llvm::size(op.views()))
-    return op.emitOpError("expected exactly ") << nViews << " view operands";
+  auto nInputsAndOutputBuffers = op.getNumInputsAndOutputBuffers();
+  if (nInputsAndOutputBuffers != llvm::size(op.views()))
+    return op.emitOpError("expected exactly ")
+           << nInputsAndOutputBuffers
+           << " inputs (tensor or buffer) and output buffer operands";
 
   auto &region = op.region();
   auto funOp = op.getFunction();
   auto funType = funOp ? funOp.getType() : FunctionType();
   if (!region.empty()) {
     if (region.getBlocks().size() != 1)
       return op.emitOpError("expected region with 1 block");
     if (failed(verifyBlockArgs(op, region.getBlocks().front())))
       return failure();
   } else {
     if (!funOp || !funOp.getType())
       return op.emitOpError(
-          "expected fun attribute to refer to a defined symbol");
-    if (failed(verifyFuncArgs(op, funType)))
+          "expected function attribute to refer to a defined symbol");
+    if (failed(verifyFuncArgsGeneric(op, funType)))
       return failure();
   }
 
   SmallVector<AffineMap, 4> indexingMaps;
   indexingMaps.reserve(op.indexing_maps().size());
   for (auto en : llvm::enumerate(op.indexing_maps())) {
     auto idx = en.index();
     auto m = en.value().template cast<AffineMapAttr>().getValue();
LogicalResult verifyGenericOp(GenericOpType op) {
   }
 
   auto concatMap = concatAffineMaps(indexingMaps);
   auto aggregateMap = inversePermutation(concatMap);
   if (!aggregateMap)
     return op.emitOpError("expected the concatenation of maps in indexing_map "
                           "to be invertible");
 
-  auto outputTensorTypes = op.getOutputTensorTypes();
-  if (outputTensorTypes.size() != op.getNumResults())
-    return op.emitOpError("expected #output tensor operands (")
-           << outputTensorTypes.size() << ") to match #results ("
-           << op.getNumResults() << ")";
-
-  unsigned index = 0;
-  for (auto it : llvm::zip(op.getResultTypes(), outputTensorTypes)) {
-    auto resTy = std::get<0>(it);
-    auto outOpTy = std::get<1>(it);
-    if (resTy != outOpTy)
-      return op.emitOpError("result #")
-             << index << " must be " << outOpTy << ", but got " << resTy;
-    ++index;
-  }
-
   return success();
 }
 
 static LogicalResult verify(GenericOp op) { return verifyGenericOp(op); }
 static LogicalResult verify(IndexedGenericOp op) { return verifyGenericOp(op); }
 
 //===----------------------------------------------------------------------===//
 // RangeOp
return failure(parser.parseOperandList(opInfo) ||
                  parser.parseOptionalAttrDict(result.attributes) ||
                  (!opInfo.empty() && parser.parseColonTypeList(types)) ||
                  parser.resolveOperands(opInfo, types, loc, result.operands));
 }
 
 template <typename GenericOpType>
 LogicalResult verifyYield(YieldOp op, GenericOpType genericOp) {
   // The operand number and types must match the view element types.
-  auto nOutputViews = genericOp.getNumOutputs();
-  if (op.getNumOperands() != nOutputViews)
-    return op.emitOpError("expected ")
-           << nOutputViews << " operand to match enclosing linalg.generic op";
+  auto nOutputs = genericOp.getNumOutputs();
+  if (op.getNumOperands() != nOutputs)
+    return op.emitOpError("expected number of yield values (")
+           << nOutputs << ") to match the number of operands of the enclosing "
+           << "linalg.generic op (" << op.getNumOperands() << ")";
 
-  for (unsigned i = 0; i != nOutputViews; ++i) {
+  for (unsigned i = 0; i != nOutputs; ++i) {
     auto elementType = genericOp.getOutputShapedType(i).getElementType();
     if (op.getOperand(i).getType() != elementType)
-      return op.emitOpError("type of return operand ")
-             << i << " (" << op.getOperand(i).getType()
-             << ") doesn't match view element type (" << elementType << ")";
+      return op.emitOpError("type of yield operand ")
+             << (i + 1) << " (" << op.getOperand(i).getType()
+             << ") doesn't match "
+             << "the element type of the enclosing linalg.generic op ("
+             << elementType << ")";
   }
   return success();
 }
 
 static LogicalResult verify(YieldOp op) {
   auto *parentOp = op.getParentOp();
   if (parentOp->getNumRegions() != 1 || parentOp->getRegion(0).empty())
     return op.emitOpError("expected single non-empty parent region");

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

static llvm::cl::list<unsigned> clTileSizes(
     llvm::cl::cat(clOptionsCategory));
 
 // Return a cloned version of `op` that operates on `loopRanges`, assumed to be
 // a subset of the original loop ranges of `op`.
 // This is achieved by applying the `loopToOperandRangesMaps` permutation maps
 // to the `loopRanges` in order to obtain view ranges.
 static LinalgOp cloneWithLoopRanges(OpBuilder &b, Location loc, LinalgOp op,
                                     ArrayRef<SubViewOp::Range> loopRanges) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
   auto maps = loopToOperandRangesMaps(op);
   SmallVector<Value, 8> clonedViews;
   clonedViews.reserve(op.getNumInputsAndOutputs());
   // Iterate over the inputs and outputs in order.
   // Extract the subranges from the linearized ranges.
-  SmallVector<Value, 8> ios(op.getInputsAndOutputs());
+  SmallVector<Value, 8> ios(op.getInputsAndOutputBuffers());
   for (auto en : llvm::enumerate(ios)) {
     unsigned idx = en.index();
     auto map = maps[idx];
     LLVM_DEBUG(dbgs() << "map: " << map << "\n");
     Value view = en.value();
     SmallVector<SubViewOp::Range, 4> viewRanges(map.getNumResults());
     for (auto en2 : llvm::enumerate(map.getResults())) {
       unsigned d = en2.index();

 };
 
 // Given an `op`, returns the first (`view`, `dimension`) pair that identifies
 // the loop range at `loopDepth`. The semantics of the loopToOperandRangesMaps
 // guarantees at least one such dimension is found. If multiple candidates exist
 // they must agree by construction (i.e. have the same size) and we just return
 // the first one.
 static ViewDimension getViewDefiningLoopRange(LinalgOp op, unsigned loopDepth) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
   auto maps = loopToOperandRangesMaps(op);
   // Iterate over the inputs and outputs in order.
   // Extract the subranges from the linearized ranges.
-  SmallVector<Value, 8> ios(op.getInputsAndOutputs());
+  SmallVector<Value, 8> ios(op.getInputsAndOutputBuffers());
   for (auto en : llvm::enumerate(ios)) {
     unsigned idx = en.index();
     auto map = maps[idx];
     LLVM_DEBUG(dbgs() << "getViewDefiningLoopRange I/O idx: " << idx << "\n");
     LLVM_DEBUG(dbgs() << "getViewDefiningLoopRange map: " << map << "\n");
     Value view = en.value();
     SmallVector<Value, 8> viewRanges(map.getNumResults(), nullptr);
     for (auto en2 : llvm::enumerate(map.getResults())) {
       if (loopDepth == en2.value().cast<AffineDimExpr>().getPosition()) {
         LLVM_DEBUG(dbgs() << "getViewDefiningLoopRange loopDepth: " << loopDepth
                           << "\n");
         LLVM_DEBUG(dbgs() << "getViewDefiningLoopRange view: " << view << "\n");
         return ViewDimension{view, static_cast<unsigned>(en2.index())};
       }
     }
   }
   llvm_unreachable("Expect to be able to extract a view defining loop range");
 }
 
 static LinalgOp fuse(Value producedView, LinalgOp producer, LinalgOp consumer,
                      unsigned consumerIdx, unsigned producerIdx,
                      OperationFolder *folder) {
+  assert(producer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   auto subView = dyn_cast_or_null<SubViewOp>(
       consumer.getInput(consumerIdx).getDefiningOp());
   auto slice =
       dyn_cast_or_null<SliceOp>(consumer.getInput(consumerIdx).getDefiningOp());
   assert(subView || slice);
   (void)subView;
   (void)slice;
 
static LinalgOp fuse(Value producedView, LinalgOp producer, LinalgOp consumer,
 
   return cloneWithLoopRanges(b, loc, producer, loopRanges);
 }
 
 // Encode structural fusion safety preconditions.
 // Some of these will be lifted in the future with better analysis.
 static bool isStructurallyFusableProducer(LinalgOp producer, Value consumedView,
                                           LinalgOp consumer) {
+  assert(producer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   if (producer.getNumOutputs() != 1) {
     LLVM_DEBUG(dbgs() << "\nNot structurally fusable (multi-output)");
     return false;
   }
   // Only fuse when the producer block dominates.
   DominanceInfo dom(producer.getOperation());
   if (!dom.dominates(producer.getOperation()->getBlock(),
                      consumer.getOperation()->getBlock())) {
     LLVM_DEBUG(
         dbgs()
         << "\nNot structurally fusable (producer block does not dominate)");
     return false;
   }
   return true;
 }
 
 bool mlir::linalg::isProducerLastWriteOfView(const LinalgDependenceGraph &graph,
                                              LinalgOp consumer,
                                              Value consumedView,
                                              LinalgOp producer) {
+  assert(producer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   // Make some simple structural checks that alleviate the need for more
   // complex analyses.
   if (!isStructurallyFusableProducer(producer, consumedView, consumer)) {
     LLVM_DEBUG(dbgs() << "\n***Not static last write due to structure:\t"
                       << *producer.getOperation());
     return false;
   }
   // Check for any interleaved write to consumedView.
   if (!graph.findCoveringWrites(producer, consumer, consumedView).empty()) {
     LLVM_DEBUG(dbgs() << "\n***Not fusable due to interleaved write:\t"
                       << *producer.getOperation());
     return false;
   }
   return true;
 }
 
 bool mlir::linalg::isFusableInto(const LinalgDependenceGraph &graph,
                                  LinalgOp consumer, Value consumedView,
                                  LinalgOp producer) {
+  assert(producer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   if (!isProducerLastWriteOfView(graph, consumer, consumedView, producer))
     return false;
   // Check for any fusion-preventing dependence to any view read/written that
   // would violate dependences.
   if (!graph.findCoveringDependences(producer, consumer).empty()) {
     LLVM_DEBUG(dbgs() << "\n***Not fusable due to an interleaved dependence:\t"
                       << *producer.getOperation());
     return false;
   }
   return true;
 }
 
 // Only consider RAW atm.
 Optional<FusionInfo> mlir::linalg::fuseProducerOf(
     OpBuilder &b, LinalgOp consumer, unsigned consumerIdx,
     const LinalgDependenceGraph &graph, OperationFolder *folder) {
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   LLVM_DEBUG(dbgs() << "\nStart examining consumer: "
                     << *consumer.getOperation());
   for (auto dependence : graph.getDependencesInto(
            consumer, LinalgDependenceGraph::DependenceType::RAW)) {
     LLVM_DEBUG(dbgs() << "\n***Consider producer:\t"
                       << *dependence.dependentOpView.op << "\n");
     auto producer = cast<LinalgOp>(dependence.dependentOpView.op);
 
     // Check that the dependence is indeed on the input `consumerIdx` view.
     auto consumedView = dependence.indexingView;
     if (consumer.getInput(consumerIdx) != consumedView)
       continue;
 
     // Consumer consumes this view, `isStructurallyFusableProducer` also checks
     // whether it is a strict subview of the producer view.
     auto producedView = dependence.dependentOpView.view;
-    auto producerIdx = producer.getIndexOfOutput(producedView).getValue();
+    auto producerIdx = producer.getIndexOfOutputBuffer(producedView).getValue();
     // `consumerIdx` and `producerIdx` exist by construction.
     LLVM_DEBUG(dbgs() << "\nRAW producer: " << *producer.getOperation()
                       << " view: " << producedView
                       << " output index: " << producerIdx);
 
     // Must be a subview or a slice to guarantee there are loops we can fuse
     // into.
     auto subView = dyn_cast_or_null<SubViewOp>(consumedView.getDefiningOp());
static void fuseLinalgOpsGreedily(FuncOp f) {
   LLVM_DEBUG(f.print(dbgs() << "\nBefore linalg-fusion: \n"));
 
   OpBuilder b(f);
   OperationFolder folder(f.getContext());
   DenseSet<Operation *> eraseSet;
 
   // Save original Linalg ops, we only want to make a pass over those.
   SmallVector<Operation *, 8> linalgOps;
-  f.walk([&](LinalgOp op) { linalgOps.push_back(op); });
+  f.walk([&](LinalgOp op) {
+    if (op.hasBufferSemantics())
+      linalgOps.push_back(op);
+  });
 
   Aliases aliases;
   LinalgDependenceGraph G(aliases, linalgOps);
   for (auto *op : llvm::reverse(linalgOps)) {
     for (unsigned consumerIdx = 0, e = LinalgOp(op).getNumInputs();
          consumerIdx < e; ++consumerIdx) {
       if (auto fusionInfo = fuseProducerOf(b, op, consumerIdx, G, &folder))
         eraseSet.insert(fusionInfo->originalProducer.getOperation());

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

 }
 
 template <typename IndexedValueType, typename LinalgOpType>
 class LinalgScopedEmitter {};
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, CopyOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs, CopyOp copyOp) {
+    assert(copyOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     auto nPar = copyOp.getNumParallelLoops();
     assert(nPar == allIvs.size());
     auto inputIvs =
         permuteIvs(allIvs.take_front(nPar), copyOp.inputPermutation());
     auto outputIvs =
         permuteIvs(allIvs.take_front(nPar), copyOp.outputPermutation());
     SmallVector<IndexHandle, 8> iivs(inputIvs.begin(), inputIvs.end());
     SmallVector<IndexHandle, 8> oivs(outputIvs.begin(), outputIvs.end());
-    IndexedValueType O(copyOp.getOutput(0)), I(copyOp.getInput(0));
+    IndexedValueType O(copyOp.getOutputBuffer(0)), I(copyOp.getInput(0));
     // Emit the proper scalar assignment, whether we are dealing with a 0-D or
     // an n-D loop nest; with or without permutations.
     // clang-format off
     nPar > 0 ? O(oivs) = I(iivs) :
                O() = I();
     // clang-format on
   }
 };
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, FillOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs, FillOp fillOp) {
+    assert(fillOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     auto nPar = fillOp.getNumParallelLoops();
     assert(nPar == allIvs.size());
     auto ivs =
         SmallVector<IndexHandle, 4>(allIvs.begin(), allIvs.begin() + nPar);
-    IndexedValueType O(fillOp.getOutput(0));
+    IndexedValueType O(fillOp.getOutputBuffer(0));
     // Emit the proper scalar assignment, whether we are dealing with a 0-D or
     // an n-D loop nest; with or without permutations.
     nPar > 0 ? O(ivs) = ValueHandle(fillOp.value())
              : O() = ValueHandle(fillOp.value());
   }
 };
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, DotOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs, DotOp dotOp) {
+    assert(dotOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     assert(allIvs.size() == 1);
     IndexHandle r_i(allIvs[0]);
     IndexedValueType A(dotOp.getInput(0)), B(dotOp.getInput(1)),
-        C(dotOp.getOutput(0));
+        C(dotOp.getOutputBuffer(0));
     // Emit scalar form.
     C() = C() + A(r_i) * B(r_i);
   }
 };
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, MatvecOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs,
                                        MatvecOp matvecOp) {
+    assert(matvecOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     assert(allIvs.size() == 2);
     IndexHandle i(allIvs[0]), r_j(allIvs[1]);
     IndexedValueType A(matvecOp.getInput(0)), B(matvecOp.getInput(1)),
-        C(matvecOp.getOutput(0));
+        C(matvecOp.getOutputBuffer(0));
     // Emit scalar form.
     C(i) = C(i) + A(i, r_j) * B(r_j);
   }
 };
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, MatmulOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs,
                                        MatmulOp matmulOp) {
+    assert(matmulOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     assert(allIvs.size() == 3);
     IndexHandle i(allIvs[0]), j(allIvs[1]), r_k(allIvs[2]);
     IndexedValueType A(matmulOp.getInput(0)), B(matmulOp.getInput(1)),
-        C(matmulOp.getOutput(0));
+        C(matmulOp.getOutputBuffer(0));
     // Emit scalar form.
     C(i, j) = C(i, j) + A(i, r_k) * B(r_k, j);
   }
 };
 
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, ConvOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs, ConvOp convOp) {
+    assert(convOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     auto b = ScopedContext::getBuilder();
     auto loc = ScopedContext::getLocation();
     auto maps = loopToOperandRangesMaps(convOp);
     SmallVector<ValueHandle, 8> fIdx(
         makeCanonicalAffineApplies(b, loc, maps[0], allIvs));
     SmallVector<ValueHandle, 8> imIdx(
         makeCanonicalAffineApplies(b, loc, maps[1], allIvs));
     SmallVector<ValueHandle, 8> oIdx(

 //            memref<?x?x?Xf32, stride_specification>
 //          store %14#1, %arg2[%i, %k, %j] :
 //            memref<?x?x?Xf32, stride_specification>
 //       }
 //      }
 //    }
 // ```
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, GenericOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs,
                                        GenericOp genericOp) {
+    assert(genericOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     auto b = ScopedContext::getBuilder();
     auto loc = ScopedContext::getLocation();
     using edsc::intrinsics::detail::ValueHandleArray;
     unsigned nInputs = genericOp.getNumInputs();
     unsigned nOutputs = genericOp.getNumOutputs();
     SmallVector<Value, 4> indexedValues(nInputs + nOutputs);
 
     // 1.a. Emit std_load from input views.
     for (unsigned i = 0; i < nInputs; ++i) {
       ValueHandleArray indexing(makeCanonicalAffineApplies(
           b, loc, genericOp.getInputIndexingMap(i), allIvs));
       indexedValues[i] = std_load(genericOp.getInput(i), indexing);
     }
 
     // 1.b. Emit std_load from output views.
     for (unsigned i = 0; i < nOutputs; ++i) {
       ValueHandleArray indexing(makeCanonicalAffineApplies(
           b, loc, genericOp.getOutputIndexingMap(i), allIvs));
-      indexedValues[nInputs + i] = std_load(genericOp.getOutput(i), indexing);
+      indexedValues[nInputs + i] =
+          std_load(genericOp.getOutputBuffer(i), indexing);
     }
 
     auto funcOp = genericOp.getFunction();
     if (funcOp) {
       // 2. Emit call.
       Operation *callOp = call(funcOp, indexedValues);
       assert(callOp->getNumResults() == genericOp.getNumOutputs());
 
       // 3. Emit std_store.
       for (unsigned i = 0; i < nOutputs; ++i) {
         ValueHandleArray indexing(makeCanonicalAffineApplies(
             b, loc, genericOp.getOutputIndexingMap(i), allIvs));
-        std_store(callOp->getResult(i), genericOp.getOutput(i), indexing);
+        std_store(callOp->getResult(i), genericOp.getOutputBuffer(i), indexing);
       }
       return;
     }
     // TODO(ntv): When a region inliner exists, use it.
     // 2. Inline region, currently only works for a single basic block.
     BlockAndValueMapping map;
     auto &block = genericOp.region().front();
     for (auto it : llvm::zip(block.getArguments(), indexedValues))
       map.map(std::get<0>(it), std::get<1>(it));
     for (auto &op : block.without_terminator()) {
       assert(op.getNumRegions() == 0);
       auto *newOp = b.clone(op, map);
       for (auto it : llvm::zip(op.getResults(), newOp->getResults()))
         map.map(std::get<0>(it), std::get<1>(it));
     }
 
     // 3. Emit std_store.
     auto *yieldOp = cast<YieldOp>(block.back()).getOperation();
     assert(yieldOp->getNumOperands() == nOutputs);
     for (unsigned i = 0; i < nOutputs; ++i) {
       ValueHandleArray indexing(makeCanonicalAffineApplies(
           b, loc, genericOp.getOutputIndexingMap(i), allIvs));
-      std_store(map.lookup(yieldOp->getOperand(i)), genericOp.getOutput(i),
-                indexing);
+      std_store(map.lookup(yieldOp->getOperand(i)),
+                genericOp.getOutputBuffer(i), indexing);
     }
   }
 };
 
 // Emits the MLIR for the scalar part of the indexed generic op by:
 //   1. Emitting std_load and std_store ops for each input and output view in
 //      order. This is achieved by applying the appropriate input or output map
 //      to the enclosing induction variables.

 //      }
 //    }
 // ```
 template <typename IndexedValueType>
 class LinalgScopedEmitter<IndexedValueType, IndexedGenericOp> {
 public:
   static void emitScalarImplementation(ArrayRef<Value> allIvs,
                                        IndexedGenericOp indexedGenericOp) {
+    assert(indexedGenericOp.hasBufferSemantics() &&
+           "expected linalg op with buffer semantics");
     auto b = ScopedContext::getBuilder();
     auto loc = ScopedContext::getLocation();
     using edsc::intrinsics::detail::ValueHandleArray;
     unsigned nInputs = indexedGenericOp.getNumInputs();
     unsigned nOutputs = indexedGenericOp.getNumOutputs();
     unsigned nLoops = allIvs.size();
     SmallVector<Value, 4> indexedValues(nLoops + nInputs + nOutputs);
 
for (unsigned i = 0; i < nInputs; ++i) {
           std_load(indexedGenericOp.getInput(i), indexing);
     }
 
     // 1.b. Emit std_load from output views.
     for (unsigned i = 0; i < nOutputs; ++i) {
       ValueHandleArray indexing(makeCanonicalAffineApplies(
           b, loc, indexedGenericOp.getOutputIndexingMap(i), allIvs));
       indexedValues[nLoops + nInputs + i] =
-          std_load(indexedGenericOp.getOutput(i), indexing);
+          std_load(indexedGenericOp.getOutputBuffer(i), indexing);
     }
 
     if (auto funcOp = indexedGenericOp.getFunction()) {
       // 2. Emit call.
       Operation *callOp = call(funcOp, indexedValues);
       assert(callOp->getNumResults() == indexedGenericOp.getNumOutputs());
 
       // 3. Emit std_store.
       for (unsigned i = 0; i < nOutputs; ++i) {
         ValueHandleArray indexing(makeCanonicalAffineApplies(
             b, loc, indexedGenericOp.getOutputIndexingMap(i), allIvs));
-        std_store(callOp->getResult(i), indexedGenericOp.getOutput(i),
+        std_store(callOp->getResult(i), indexedGenericOp.getOutputBuffer(i),
                   indexing);
       }
       return;
     }
     // TODO(ntv): When a region inliner exists, use it.
     // 2. Inline region, currently only works for a single basic block.
     BlockAndValueMapping map;
     auto &block = indexedGenericOp.region().front();
     for (auto it : llvm::zip(block.getArguments(), indexedValues))
       map.map(std::get<0>(it), std::get<1>(it));
     for (auto &op : block.without_terminator()) {
       assert(op.getNumRegions() == 0);
       auto *newOp = b.clone(op, map);
       for (auto it : llvm::zip(op.getResults(), newOp->getResults()))
         map.map(std::get<0>(it), std::get<1>(it));
     }
 
     // 3. Emit std_store.
     auto *yieldOp = cast<YieldOp>(block.back()).getOperation();
     assert(yieldOp->getNumOperands() == nOutputs);
     for (unsigned i = 0; i < nOutputs; ++i) {
       ValueHandleArray indexing(makeCanonicalAffineApplies(
           b, loc, indexedGenericOp.getOutputIndexingMap(i), allIvs));
       std_store(map.lookup(yieldOp->getOperand(i)),
-                indexedGenericOp.getOutput(i), indexing);
+                indexedGenericOp.getOutputBuffer(i), indexing);
     }
   }
 };
 
 namespace {
 // This struct is for factoring out the implementation and support template
 // instantiations in the following 2 cases:
 //   1. Appending to a list of patterns via RewritePatternList.

 LogicalResult LinalgOpToLoopsImpl<LoopTy, IndexedValueTy, ConcreteOpTy>::doit(
     Operation *op, PatternRewriter &rewriter) {
   OpBuilder b(op);
   ScopedContext scope(b, op->getLoc());
 
   // The flattened loopToOperandRangesMaps is expected to be an invertible
   // permutation map (which is asserted in the inverse calculation).
   auto linalgOp = cast<ConcreteOpTy>(op);
+  assert(linalgOp.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   auto invertedMap =
       inversePermutation(concatAffineMaps(loopToOperandRangesMaps(linalgOp)));
   if (!invertedMap) {
     LinalgScopedEmitter<IndexedValueTy, ConcreteOpTy>::emitScalarImplementation(
         {}, linalgOp);
     return success();
   }
 

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

for (auto *originalProducer : originalProducers)
     rewriter.eraseOp(originalProducer);
   return success();
 }
 
 bool mlir::linalg::detail::isProducedByOpOfTypeImpl(
     Operation *consumerOp, Value consumedView,
     function_ref<bool(Operation *)> isaOpType) {
   LinalgOp consumer = dyn_cast<LinalgOp>(consumerOp);
+  assert(consumer.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   if (!consumer)
     return false;
 
   auto maybeConsumerIndex = consumer.getIndexOfInput(consumedView);
   if (!maybeConsumerIndex)
     return false;
 
   Aliases aliases;
mlir::linalg::vectorizeGenericLinalgOpPrecondition(Operation *op) {
 
   // TODO(ntv): non-identity layout.
   auto isStaticMemRefWithIdentityLayout = [](Value v) {
     auto m = v.getType().dyn_cast<MemRefType>();
     if (!m || !m.hasStaticShape() || !m.getAffineMaps().empty())
       return false;
     return true;
   };
-  if (!llvm::all_of(genericOp.getInputsAndOutputs(),
+  if (!llvm::all_of(genericOp.getInputsAndOutputBuffers(),
                     isStaticMemRefWithIdentityLayout))
     return failure();
   return success();
 }
 
 SmallVector<Value, 0>
 mlir::linalg::vectorizeGenericLinalgOp(PatternRewriter &rewriter,
                                        Operation *op) {
   LLVM_DEBUG(dbgs() << "\n[" DEBUG_TYPE
                        "]: Rewrite linalg op as vector.contract: "
                     << *op << ":\n");
 
   assert(succeeded(vectorizeGenericLinalgOpPrecondition(op)) &&
          "DRR failure case must be a precondition");
 
   auto genericOp = cast<linalg::GenericOp>(op);
+  assert(genericOp.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   edsc::ScopedContext scope(rewriter, op->getLoc());
   using edsc::intrinsics::std_load;
   using edsc::intrinsics::std_store;
   using vector_contract = edsc::intrinsics::ValueBuilder<vector::ContractionOp>;
   using vector_type_cast = edsc::intrinsics::ValueBuilder<vector::TypeCastOp>;
   auto vA = std_load(vector_type_cast(genericOp.getInput(0)));
   auto vB = std_load(vector_type_cast(genericOp.getInput(1)));
-  auto vectorMemRefC = vector_type_cast(genericOp.getOutput(0));
+  auto vectorMemRefC = vector_type_cast(genericOp.getOutputBuffer(0));
   auto vC = std_load(vectorMemRefC);
   auto vRes = vector_contract(vA, vB, vC, genericOp.indexing_maps(),
                               genericOp.iterator_types());
   std_store(vRes, vectorMemRefC);
   return {};
 }
 
 //============================================================================//

 //============================================================================//
 // Precondition and transformation for Linalg subview promotion.
 //============================================================================//
 LogicalResult mlir::linalg::promoteSubviewsLinalgOpPrecondition(Operation *op) {
   LinalgOp linOp = dyn_cast<LinalgOp>(op);
   // Transformation applies to buffers only.
   if (!linOp || !linOp.hasBufferSemantics())
     return failure();
-  if (llvm::none_of(linOp.getInputsAndOutputs(), [](Value v) {
+  if (llvm::none_of(linOp.getInputsAndOutputBuffers(), [](Value v) {
         return isa_and_nonnull<SubViewOp>(v.getDefiningOp());
       }))
     return failure();
   return success();
 }
 
 SmallVector<Value, 0>
 mlir::linalg::promoteSubviewsLinalgOp(PatternRewriter &rewriter,
                                       Operation *op) {
   LLVM_DEBUG(dbgs() << "\n[" DEBUG_TYPE "]: Promote subviews for linalg op: "
                     << *op << ":\n");
 
   assert(succeeded(promoteSubviewsLinalgOpPrecondition(op)) &&
          "DRR failure case must be a precondition");
 
   LinalgOp linOp = cast<LinalgOp>(op);
+  assert(linOp.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   SetVector<Value> subViews;
-  for (auto it : linOp.getInputsAndOutputs())
+  for (auto it : linOp.getInputsAndOutputBuffers())
     if (auto sv = dyn_cast_or_null<SubViewOp>(it.getDefiningOp()))
       subViews.insert(sv);
   if (!subViews.empty()) {
     promoteSubViewOperands(rewriter, linOp, subViews);
     return {};
   }
   llvm_unreachable("DRR failure case must be a precondition");
 }

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

mlir::linalg::promoteSubViews(OpBuilder &b, Location loc,
   }
   return res;
 }
 
 LinalgOp mlir::linalg::promoteSubViewOperands(OpBuilder &b, LinalgOp op,
                                               SetVector<Value> subViews,
                                               bool dynamicBuffers,
                                               OperationFolder *folder) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
+
   // 1. Promote the specified views and use them in the new op.
   ScopedContext scope(b, op.getLoc());
   auto promotedBufferAndViews = promoteSubViews(
       b, op.getLoc(), subViews.getArrayRef(), dynamicBuffers, folder);
   SmallVector<Value, 8> opViews;
   opViews.reserve(op.getNumInputsAndOutputs());
   SmallVector<std::pair<Value, Value>, 8> writebackViews;
   writebackViews.reserve(subViews.size());
   unsigned promotedIdx = 0;
-  for (auto view : op.getInputsAndOutputs()) {
+  for (auto view : op.getInputsAndOutputBuffers()) {
     if (subViews.count(view) != 0) {
       opViews.push_back(promotedBufferAndViews[promotedIdx].fullLocalView);
       writebackViews.emplace_back(std::make_pair(
           view, promotedBufferAndViews[promotedIdx].partialLocalView));
       promotedIdx++;
     } else {
       opViews.push_back(view);
     }
   }
 
   // 2. Append all other operands as they appear, this enforces that such
   // operands are not views. This is to support cases such as FillOp taking
   // extra scalars etc.
   auto operands = getAssumedNonViewOperands(op);
   opViews.append(operands.begin(), operands.end());
   LinalgOp res = op.clone(b, op.getLoc(), opViews);
 
   // 3. Emit write-back for the promoted output views: copy the partial view.
   for (auto viewAndPartialLocalView : writebackViews) {
     // WARNING: MUST use the old op to determine whether the operand view is an
     // output.
     bool isOutput =
-        op.getIndexOfOutput(viewAndPartialLocalView.first).hasValue();
+        op.getIndexOfOutputBuffer(viewAndPartialLocalView.first).hasValue();
     if (isOutput)
       copy(viewAndPartialLocalView.second, viewAndPartialLocalView.first);
   }
 
   // 4. Dealloc local buffers.
   for (const auto &pi : promotedBufferAndViews)
     dealloc(pi.buffer);
 
   return res;
 }
 
 static void promoteSubViews(FuncOp f, bool dynamicBuffers) {
   SmallVector<LinalgOp, 8> toErase;
   OperationFolder folder(f.getContext());
   f.walk([dynamicBuffers, &folder, &toErase](LinalgOp op) {
+    if (!op.hasBufferSemantics())
+      return;
+
     // TODO(ntv) some heuristic here to decide what to promote. Atm it is all or
     // nothing.
     SetVector<Value> subViews;
     OpBuilder b(op);
-    for (auto it : op.getInputsAndOutputs())
+    for (auto it : op.getInputsAndOutputBuffers())
       if (auto sv = dyn_cast_or_null<SubViewOp>(it.getDefiningOp()))
         subViews.insert(sv);
     if (!subViews.empty()) {
       promoteSubViewOperands(b, op, subViews, dynamicBuffers, &folder);
       toErase.push_back(op);
     }
   });
   for (auto op : toErase)

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

 //   }
 // }
 //
 // TODO(pifon, ntv): Investigate whether mixing implicit and explicit indices
 // does not lead to losing information.
 void transformIndexedGenericOpIndices(
     OpBuilder &b, LinalgOp op, ArrayRef<ValueHandle *> pivs,
     const LoopIndexToRangeIndexMap &loopIndexToRangeIndex) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
   auto indexedGenericOp = dyn_cast<IndexedGenericOp>(op.getOperation());
   if (!indexedGenericOp)
     return;
 
   // `linalg.indexed_generic` comes in two flavours. One has a region with a
   // single block that defines the loop body. The other has a `fun` attribute
   // that refers to an existing function symbol. The `fun` function call will be
   // inserted in the loop body in that case.
if (isTiled(map.getResult(r), tileSizes))
       return true;
   return false;
 }
 
 static SmallVector<Value, 4>
 makeTiledViews(OpBuilder &b, Location loc, LinalgOp linalgOp,
                ArrayRef<Value> ivs, ArrayRef<Value> tileSizes,
                ArrayRef<Value> viewSizes, OperationFolder *folder) {
+  assert(linalgOp.hasBufferSemantics() &&
+         "expected linalg op with buffer semantics");
   assert(ivs.size() == static_cast<size_t>(llvm::count_if(
                            llvm::make_range(tileSizes.begin(), tileSizes.end()),
                            [](Value v) { return !isZero(v); })) &&
          "expected as many ivs as non-zero sizes");
 
   using edsc::intrinsics::select;
   using edsc::op::operator+;
   using edsc::op::operator<;
 
   // Construct (potentially temporary) mins and maxes on which to apply maps
   // that define tile subviews.
   SmallVector<Value, 8> lbs, subViewSizes;
   for (unsigned idx = 0, idxIvs = 0, e = tileSizes.size(); idx < e; ++idx) {
     bool isTiled = !isZero(tileSizes[idx]);
     lbs.push_back(isTiled ? ivs[idxIvs++] : (Value)constant_index(folder, 0));
     subViewSizes.push_back(isTiled ? tileSizes[idx] : viewSizes[idx]);
   }
 
   auto *op = linalgOp.getOperation();
 
   SmallVector<Value, 4> res;
   res.reserve(op->getNumOperands());
-  auto viewIteratorBegin = linalgOp.getInputsAndOutputs().begin();
+  auto viewIteratorBegin = linalgOp.getInputsAndOutputBuffers().begin();
   for (unsigned viewIndex = 0; viewIndex < linalgOp.getNumInputsAndOutputs();
        ++viewIndex) {
     Value view = *(viewIteratorBegin + viewIndex);
     unsigned rank = view.getType().cast<MemRefType>().getRank();
     auto map = loopToOperandRangesMaps(linalgOp)[viewIndex];
     // If the view is not tiled, we can use it as is.
     if (!isTiled(map, tileSizes)) {
       res.push_back(view);
makeTiledViews(OpBuilder &b, Location loc, LinalgOp linalgOp,
 
   return res;
 }
 
 Optional<TiledLinalgOp>
 mlir::linalg::tileLinalgOp(OpBuilder &b, LinalgOp op, ArrayRef<Value> tileSizes,
                            ArrayRef<unsigned> permutation,
                            OperationFolder *folder) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
   // 1. Enforce the convention that "tiling by zero" skips tiling a particular
   // dimension. This convention is significantly simpler to handle instead of
   // adjusting affine maps to account for missing dimensions.
   assert(op.getNumParallelLoops() + op.getNumReductionLoops() +
                  op.getNumWindowLoops() ==
              tileSizes.size() &&
          "expected matching number of tile sizes and loops");
 
for (auto iv : ivs)
     loops.push_back(loop::getForInductionVarOwner(iv));
 
   return TiledLinalgOp{res, loops};
 }
 
 Optional<TiledLinalgOp> mlir::linalg::tileLinalgOp(
     OpBuilder &b, LinalgOp op, ArrayRef<int64_t> tileSizes,
     ArrayRef<unsigned> permutation, OperationFolder *folder) {
+  assert(op.hasBufferSemantics() && "expected linalg op with buffer semantics");
   if (tileSizes.empty())
     return llvm::None;
 
   // The following uses the convention that "tiling by zero" skips tiling a
   // particular dimension. This convention is significantly simpler to handle
   // instead of adjusting affine maps to account for missing dimensions.
   auto nLoops = op.getNumParallelLoops() + op.getNumReductionLoops() +
                 op.getNumWindowLoops();
Optional<TiledLinalgOp> mlir::linalg::tileLinalgOp(
 
   return tileLinalgOp(b, op, tileSizeValues, permutation, folder);
 }
 
 static void tileLinalgOps(FuncOp f, ArrayRef<int64_t> tileSizes) {
   OpBuilder b(f);
   OperationFolder folder(f.getContext());
   f.walk([tileSizes, &b, &folder](LinalgOp op) {
+    if (!op.hasBufferSemantics())
+      return;
     auto opLoopsPair =
         tileLinalgOp(b, op, tileSizes, /*permutation=*/{}, &folder);
     // If tiling occurred successfully, erase old op.
     if (opLoopsPair)
       op.erase();
   });
   f.walk([](LinalgOp op) {
     if (!op.getOperation()->hasNoSideEffect())

mlir/test/Dialect/Linalg/invalid.mlir

linalg.generic {
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0: memref<f32>
 }
 
 // -----
 
 func @generic_exactly_2_views(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected exactly 2 view operands}}
+  // expected-error @+1 {{op expected exactly 2 inputs (tensor or buffer) and output buffer operands}}
   linalg.generic {
     args_in = 1,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0, %arg0, %arg0: memref<f32>, memref<f32>, memref<f32>
 }
 
 // -----
 
 func @generic_undefined_fun(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun attribute to refer to a defined symbol}}
+  // expected-error @+1 {{op expected function attribute to refer to a defined symbol}}
   linalg.generic {
     args_in = 1,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0, %arg0: memref<f32>, memref<f32>
 }
 
 // -----
 
 func @foo() { return }
 
 func @generic_mismatched_num_arguments(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun arguments to match number of views}}
+  // expected-error @+1 {{op expected function arguments to match number of operands}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0: memref<f32>
 }
 
 // -----
 
 func @foo(%0: i32) { return }
 
 func @generic_mismatched_num_returns(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun results to match number of output views}}
+  // expected-error @+1 {{op expected function results(0) to match number of outputs(1)}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0: memref<f32>
 }
 
 // -----
 
+func @foo(%0: i32, %1: i32, %2: i32) { return }
+
+func @generic_mismatched_num_returns(%0: memref<i32>, %1: memref<f32>) {
+  // expected-error @+1 {{op expected function argument 2 of the same type as elemental type 'f32' of operand 2}}
+  linalg.generic {
+    args_in = 3,
+    args_out = 0,
+    fun = @foo,
+    indexing_maps =  [ () -> (0) ],
+    iterator_types = []
+  } %0, %1, %1: memref<i32>, memref<f32>, memref<f32>
+}
+
+// -----
+
+func @foo(%0: i32, %1: i32, %2: f32) -> i32 { return %1: i32}
+
+func @generic_mismatched_num_returns(%0: memref<i32>, %1: memref<f32>) {
+  // expected-error @+1 {{op expected function result 1 of the same type as elemental type 'f32' of output 1}}
+  linalg.generic {
+    args_in = 2,
+    args_out = 1,
+    fun = @foo,
+    indexing_maps =  [ () -> (0) ],
+    iterator_types = []
+  } %0, %0, %1: memref<i32>, memref<i32>, memref<f32>
+}
+
+// -----
+
 func @foo(%0: i32) -> i32 { return %0: i32 }
 
 func @generic_symbol_in_map(%arg0: memref<i32>) {
   // expected-error @+1 {{op expected indexing_map #0 to have no symbols}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     fun = @foo,

 // -----
 
 func @foo(%0: i32) -> f32 {
   %1 = constant 0.0: f32
   return %1: f32
 }
 
 func @generic_fun_arg_0_element_type(%arg0: memref<?xf32, (i)[off]->(off + i)>) {
-  // expected-error @+1 {{op expected fun argument 0 of the same type as elemental type 'f32' of view 0}}
+  // expected-error @+1 {{op expected function argument 1 of the same type as elemental type 'f32' of operand 1}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0: memref<?xf32, (i)[off]->(off + i)>
 }
 
 // -----
 
 func @foo(%0: f32) -> i4 {
   %1 = constant 1: i4
   return %1: i4
 }
 
 func @generic_fun_result_0_element_type(%arg0: memref<?xf32, (i)[off]->(off + i)>) {
-  // expected-error @+1 {{op expected fun result 0 of the same type as elemental type 'f32' of view 0}}
+  // expected-error @+1 {{op expected function result 1 of the same type as elemental type 'f32' of output 1}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     fun = @foo,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0: memref<?xf32, (i)[off]->(off + i)>
 }
linalg.generic {
     ^bb1:
     ^bb2:
   }: memref<f32>, memref<f32>
 }
 
 // -----
 
 func @generic_mismatched_num_arguments(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected number of block arguments to match number of views}}
+  // expected-error @+1 {{op expected number of block arguments to match number of operands}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0 {
     ^bb:
   }: memref<f32>
 }
 
 // -----
 
 func @generic_block_arg_type(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected block argument 0 of the same type as elemental type of output view: 'memref<f32>'}}
+  // expected-error @+1 {{op expected block argument 1 of the same type as elemental type of output operand: 'memref<f32>'}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ () -> (0) ],
     iterator_types = []
   } %arg0 {
     ^bb(%i: i1):
   }: memref<f32>
 }
 
 // -----
 
 func @indexed_generic_block_arg_count(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected number of block arguments to match number of views + number of loops}}
+  // expected-error @+1 {{op expected number of block arguments to match number of operands + number of loops}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"]
   } %arg0 {
     ^bb(%f: f32):
   }: memref<f32>
 }
 
 // -----
 
 func @indexed_generic_block_induction_var_arg_type(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected block argument 0 to be of IndexType}}
+  // expected-error @+1 {{op expected block argument 1 to be an index}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"]
   } %arg0 {
     ^bb(%i: f64, %f: f32):
   }: memref<f32>
 }
 
 // -----
 
 func @indexed_generic_block_arg_type(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected block argument 1 of the same type as elemental type of output view: 'memref<f32>'}}
+  // expected-error @+1 {{op expected block argument 2 of the same type as elemental type of output operand: 'memref<f32>'}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"]
   } %arg0 {
     ^bb(%i: index, %f: i1):
   }: memref<f32>
 }
 
 // -----
 
 func @foo(%f: f32) -> (f32) {
   return %f : f32
 }
 func @indexed_generic_fun_arg_count(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun arguments to match number of views + number of loops}}
+  // expected-error @+1 {{op expected function arguments to match number of loops + number of operands}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"],
     fun = @foo
   } %arg0:  memref<f32>
 }
 
 // -----
 
 func @foo(%i: i32, %val: f32) -> (f32) {
   return %val : f32
 }
 func @indexed_generic_fun_induction_var_arg_type(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun argument 0 to be of IndexType}}
+  // expected-error @+1 {{op expected function argument 1 to be an index}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     iterator_types = ["parallel"],
     indexing_maps = [ (i) -> (i) ],
     fun = @foo
   } %arg0 : memref<f32>
 }
 
 // -----
 
 func @foo(%i: index, %val: i1) -> (i1) {
   return %val : i1
 }
 func @indexed_generic_fun_arg_type(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun argument 1 of the same type as elemental type 'f32' of view 0}}
+  // expected-error @+1 {{op expected function argument 2 of the same type as elemental type 'f32' of input 1}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"],
     fun = @foo
   } %arg0: memref<f32>
 }
 
 // -----
 
 func @foo(%i: index, %val: i1) -> (i1, i1) {
   return %val, %val : i1, i1
 }
 func @indexed_generic_fun_result_count(%arg0: memref<f32>) {
-  // expected-error @+1 {{op expected fun results to match number of output views}}
+  // expected-error @+1 {{op expected function results to match number of outputs}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"],
     fun = @foo
   } %arg0: memref<f32>
 }
 
 // -----
 
 func @foo(%i: index, %val: i32) -> (f32) {
   %val_float = sitofp %val : i32 to f32
   return %val_float : f32
 }
 func @indexed_generic_fun_result_count(%arg0: memref<i32>) {
-  // expected-error @+1 {{op expected fun result 0 of the same type as elemental type 'i32' of view 0}}
+  // expected-error @+1 {{op expected function result 1 of the same type as elemental type 'i32' of output 1}}
   linalg.indexed_generic {
     args_in = 0,
     args_out = 1,
     indexing_maps =  [ (d0) -> (d0) ],
     iterator_types = ["parallel"],
     fun = @foo
   } %arg0: memref<i32>
 }
 
 // -----
 
 func @generic_fun_result_0_element_type(%arg0: memref<?xf32, (i)[off]->(off + i)>) {
-  // expected-error @+9 {{type of return operand 0 ('i1') doesn't match view element type ('f32')}}
+  // expected-error @+9 {{type of yield operand 1 ('i1') doesn't match the element type of the enclosing linalg.generic op ('f32')}}
   linalg.generic {
     args_in = 0,
     args_out = 1,
     indexing_maps = [ (i) -> (i) ],
     iterator_types = ["parallel"]
   } %arg0 {
     ^bb(%i: f32):
       %0 = constant 0: i1
func @generic_result_tensor_type(%arg0: memref<?xf32, (i)[off]->(off + i)>) {
   } %arg0 {
     ^bb(%i: f32):
       linalg.yield %i: f32
   }: memref<?xf32, (i)[off]->(off + i)> -> f32
 }
 
 // -----
 
-func @generic_result_tensor_count(%arg0: memref<?xf32, (i)[off]->(off + i)>) {
-  // expected-error @+1 {{op expected #output tensor operands (0) to match #results (1)}}
-  %0 = linalg.generic {
-    args_in = 0,
-    args_out = 1,
-    indexing_maps = [ (i) -> (i) ],
-    iterator_types = ["parallel"]
-  } %arg0 {
-    ^bb(%i: f32):
-      linalg.yield %i: f32
-  }: memref<?xf32, (i)[off]->(off + i)> -> tensor<?xf32>
-}
-
-// -----
-
-func @generic_result_tensor_type(%arg0: tensor<?xf32>) {
-  // expected-error @+1 {{op result #0 must be 'tensor<?xf32>', but got 'tensor<?x?xf32>'}}
-  %0 = linalg.generic {
-    args_in = 0,
-    args_out = 1,
-    indexing_maps = [ (i) -> (i) ],
-    iterator_types = ["parallel"]
-  } %arg0 {
-    ^bb(%i: f32):
-      linalg.yield %i: f32
-  }: tensor<?xf32> -> tensor<?x?xf32>
-}
-
-// -----
-
 func @generic_fun_result_0_element_type(%arg0: memref<?xf32>) {
   // expected-error @+1 {{'linalg.dot' op expected 3 or more operands}}
   linalg.dot(%arg0, %arg0): memref<?xf32>, memref<?xf32>
 }
 
 // -----
 
 // expected-error @+1 {{unknown Linalg type}}

mlir/test/Dialect/Linalg/roundtrip.mlir

 
 func @generic_with_tensor_input(%arg0: tensor<?x?xvector<3x4xi4>>, %arg1: memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>) {
   linalg.generic #trait %arg0, %arg1 {foo = 1} : tensor<?x?xvector<3x4xi4>>, memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>
   return
 }
 // CHECK-LABEL: func @generic_with_tensor_input
 //       CHECK:   linalg.generic {args_in = 1 : i64, args_out = 1 : i64, fun = @foo, indexing_maps = [#{{.*}}, #{{.*}}], iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_1"} %{{.*}}, %{{.*}} {foo = 1 : i64}: tensor<?x?xvector<3x4xi4>>, memref<?x?x?xf32, #[[strided3D]]>
 
-func @generic_with_tensor_output(%arg0: memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>, %arg1: tensor<?x?x?xf32>) -> (tensor<?x?x?xf32>) {
-  %0 = linalg.generic #trait %arg0, %arg1 {foo = 1} : memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
-  return %0 : tensor<?x?x?xf32>
+#trait2 = {
+  args_in = 2,
+  args_out = 1,
+  indexing_maps = #accesses,
+  iterator_types = ["parallel", "parallel", "parallel"],
+  fun = @foo,
+  library_call = "some_external_function_name_1"
 }
-// CHECK-LABEL: func @generic_with_tensor_output
-//       CHECK:   linalg.generic {args_in = 1 : i64, args_out = 1 : i64, fun = @foo, indexing_maps = [#{{.*}}, #{{.*}}], iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_1"} %{{.*}}, %{{.*}} {foo = 1 : i64}: memref<?x?xvector<3x4xi4>, #[[strided2D]]>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
-//       CHECK:   return {{.*}} : tensor<?x?x?xf32>
-
 func @generic_with_tensor_input_and_output(%arg0: tensor<?x?xvector<3x4xi4>>, %arg1: tensor<?x?x?xf32>) -> (tensor<?x?x?xf32>) {
-  %0 = linalg.generic #trait %arg0, %arg1 {foo = 1} : tensor<?x?xvector<3x4xi4>>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
+  %0 = linalg.generic #trait2 %arg0, %arg1 {foo = 1} : tensor<?x?xvector<3x4xi4>>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
   return %0 : tensor<?x?x?xf32>
 }
 // CHECK-LABEL: func @generic_with_tensor_input_and_output
-//       CHECK:   linalg.generic {args_in = 1 : i64, args_out = 1 : i64, fun = @foo, indexing_maps = [#{{.*}}, #{{.*}}], iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_1"} %{{.*}}, %{{.*}} {foo = 1 : i64}: tensor<?x?xvector<3x4xi4>>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
+//       CHECK:   linalg.generic {args_in = 2 : i64, args_out = 1 : i64, fun = @foo, indexing_maps = [#{{.*}}, #{{.*}}], iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_1"} %{{.*}}, %{{.*}} {foo = 1 : i64}: tensor<?x?xvector<3x4xi4>>, tensor<?x?x?xf32> -> tensor<?x?x?xf32>
 //       CHECK:   return {{.*}} : tensor<?x?x?xf32>
 
-#trait2 = {
+#trait3 = {
   args_in = 1,
   args_out = 1,
   indexing_maps = #accesses,
   iterator_types = ["parallel", "parallel", "parallel"],
   library_call = "some_external_function_name_2"
 }
 func @generic_region(%arg0: memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>, %arg1: memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>) {
-  linalg.generic #trait2 %arg0, %arg1 {
+  linalg.generic #trait3 %arg0, %arg1 {
     ^bb(%a: vector<3x4xi4>, %b: f32) :
       linalg.yield %b : f32
   } {foo = 1}: memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>, memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>
   return
 }
 // CHECK-LABEL: func @generic_region
 //       CHECK:   linalg.generic {args_in = 1 : i64, args_out = 1 : i64, indexing_maps = [#{{.*}}, #{{.*}}], iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_2"} %{{.*}}, %{{.*}} {
 //       CHECK:    ^{{.*}}(%{{.*}}: vector<3x4xi4>, %{{.*}}: f32):    // no predecessors
 //       CHECK:      linalg.yield %{{.*}} : f32
 //       CHECK:    } {foo = 1 : i64}: memref<?x?xvector<3x4xi4>, #[[strided2D]]>, memref<?x?x?xf32, #[[strided3D]]>
 func @indexed_generic(%arg0: memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>,
                       %arg1: memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>) {
-  linalg.indexed_generic #trait2 %arg0, %arg1 {
+  linalg.indexed_generic #trait3 %arg0, %arg1 {
   ^bb(%i: index, %j: index, %k: index, %a: vector<3x4xi4>, %b: f32) :
       linalg.yield %b : f32
   } {foo = 1}: memref<?x?xvector<3x4xi4>, offset: ?, strides: [?, 1]>, memref<?x?x?xf32, offset: ?, strides: [?, ?, 1]>
   return
 }
 // CHECK-LABEL: func @indexed_generic
 //       CHECK:   linalg.indexed_generic {args_in = 1 : i64, args_out = 1 : i64, indexing_maps = [#{{.*}}, #{{.*}}],  iterator_types = ["parallel", "parallel", "parallel"], library_call = "some_external_function_name_2"} %{{.*}}, %{{.*}} {
 //       CHECK:    ^{{.*}}(%{{.*}}: index, %{{.*}}: index, %{{.*}}: index, %{{.*}}: vector<3x4xi4>, %{{.*}}: f32):