[mlir] Split std.splat into tensor.splat and vector.splat

[rriddle]

Summary

This is part of the larger effort to split the standard dialect. This will also allow for pruning some
additional dependencies on Standard (done in a followup).

[mehdi_amini]

Was this part of an RFC already?

[rriddle]

In D118202#3271013, @mehdi_amini wrote:

Was this part of an RFC already?

Can't remember, let me see. (Haha if so, just getting rid of things as I see them in the codebase)

[rriddle]

In D118202#3271018, @rriddle wrote:

In D118202#3271013, @mehdi_amini wrote:

Was this part of an RFC already?

Can't remember, let me see. (Haha if so, just getting rid of things as I see them in the codebase)

Couldn't find one when searching on discourse, it was left as a TODO when the tensor dialect was originally split. I'd like to just split it now given that the standard dialect has ~12 ops left.

[bondhugula]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

This is just duplicating the entire folding hook. Move this method to lib/IR/BuiltinTypeInterfaces.cpp and reuse from both places?

[rriddle]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

Doesn't feel like enough code to warrant sharing (mostly just unnecessary asserts that are already invariant to this method), cleaned up the code a bit.

[mehdi_amini]

mlir/lib/Dialect/Vector/VectorOps.cpp

2479 ↗

(On Diff #403077)

We lost the get prefix? Aren't all dialects generating both forms by now?

[rriddle]

mlir/lib/Dialect/Vector/VectorOps.cpp

2479 ↗

(On Diff #403077)

Yeah, I was also a bit surprised. Not sure what the status of the prefix flipping is, but I can look into flipping it myself.

[ThomasRaoux]

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.
@nicolasvasilache do you have any opinion? Any reason to have both?

[bondhugula]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

Isn't the splat on a vector type functionally identical to a splat on a tensor type? Is there a need for two different splat op versions on shaped pure value types? (memref type vs tensor type isn't the same as vector type vs tensor type) The folding hook can get longer and is already missing many trivial optimizations here, which could in the future use both vector dialect ops and tensor dialect ops.

I feel a split should be motivated by the need for two different op versions as opposed to simply trimming the standard dialect. (For eg. the folding hooks for tensor.dim and memref.dim have diverged and specialized and that's suitable there, but I'm not sure that's the case here.)

[rriddle]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

From a layering perspective right now it makes no sense to keep them unified. The functionality built on top of both is also so small right now that it brings more burden than benefit for keeping them unified. I'd rather separate and layer them properly in their own respective dialects (and help achieve a valuable goal of killing the standard dialect), and then determine what can be unified as things are actually built out.

As mentioned below as well, there is also a sort of awkward duplication with std.splat and vector.broadcast.

[rriddle]

In D118202#3271122, @ThomasRaoux wrote:

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.
@nicolasvasilache do you have any opinion? Any reason to have both?

Yeah, it's kind of awkward. I've noticed that vector.broadcast lowers to splat for some of its cases, for what I would assume should be for behavior that vector.broadcast already supports. It'd be nice to clean that up.

[mehdi_amini]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

We can take it the other way around: if the standard dialect didn't exist and someone needed this operation today, how would we add it?

[bondhugula]

In D118202#3271122, @ThomasRaoux wrote:

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.

A splat is on an "elemental" type while a broadcast is on an "elemental" or a "tensor" type and thus a superset as you say. It is trivial to convert a vector.splat to a vector.broadcast if you don't want to support both for the purposes of a transformation. Ultimately, it's a splat that most commonly corresponds to hardware intrinsics since those are defined on repeating elements -- while a broadcast is a higher-order abstraction. You can always canonicalize a vector.broadcast to a vector.splat if the operand is a scalar. I am a strong -1 on removing splat on vectors -- in fact, I feel vector.broadcast could drop scalar operand support and instead be exclusively for n-d vector operands.

[jpienaar]

mlir/lib/Dialect/Vector/VectorOps.cpp

2479 ↗

(On Diff #403077)

I thought I flipped vector too ... Mmm. I may have forgotten to send it out. There is the procedure for flipping and the clang-tooling based rewrite if needed (even if I abuse clang-tidy a bit for that)

[bondhugula]

mlir/lib/Dialect/Vector/VectorOps.cpp

4278–4291 ↗

(On Diff #403060)

As mentioned below as well, there is also a sort of awkward duplication with std.splat and vector.broadcast.

That overlap is unrelated to this issue and mostly from an oversight when vector.broadcast was added: in fact, it's awkward that vector.broadcast right now broadcasts either a scalar or an n-d vector -- two very different types. Ideally, vector.broadcast should only support *vector typed* operands while splat supports scalar operands. The fact that the latter scenario is often the common case with direct hardware support (or close to that) warrants retaining such an operation -- this is aligned with the purposes of vector types and vector dialect as opposed to tensors. (Not that it's important now, in fact, when I introduced the splat op, the intent was to mainly use it only for vector types.)

[nicolasvasilache]

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.
@nicolasvasilache do you have any opinion? Any reason to have both?

Yes, dropping the antiquated splat, I don't see value in keeping it.

[rriddle]

In D118202#3271598, @nicolasvasilache wrote:

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.
@nicolasvasilache do you have any opinion? Any reason to have both?

Yes, dropping the antiquated splat, I don't see value in keeping it.

I'm happy to kill it, but it begs the question of which comes first. I have a slight preference for pushing this commit first for a couple of reasons.
Partially because it is a necessary refactoring and has some dependencies stacked on top of it. Removing splat requires untangling some interconnected things, which are a bit easier to manage when all of the code is in one dialect. Lastly, because there is some contention from Uday about having splat (that I don't want to bog down this necessary commit too much with discussing).

[nicolasvasilache]

In D118202#3271640, @rriddle wrote:

In D118202#3271598, @nicolasvasilache wrote:

The vector dialect has vector.broadcast op that is strictly a superset of splat. Personally I found it annoying to have to support both so I would prefer not having vector.splat at all.
@nicolasvasilache do you have any opinion? Any reason to have both?

Yes, dropping the antiquated splat, I don't see value in keeping it.

I'm happy to kill it, but it begs the question of which comes first. I have a slight preference for pushing this commit first for a couple of reasons.
Partially because it is a necessary refactoring and has some dependencies stacked on top of it. Removing splat requires untangling some interconnected things, which are a bit easier to manage when all of the code is in one dialect. Lastly, because there is some contention from Uday about having splat (that I don't want to bog down this necessary commit too much with discussing).

As you know I am a fan of faster iteration once we know where we are going so I'm happily giving you green light on this.

[rriddle]

Ping. Is there something inherently blocking here? This puts the world in an objectively better state than what we are now, and is kind of blocking the effort to kill the standard dialect.

[bondhugula]

LGTM for larger reasons although it's unfortunate we've created an unnecessary duplicate. There is no reason for two ops doing the ditto -- both VectorType and TensorType are both "value" shaped types and "splat" does the *same* thing on both.

Yes, dropping the antiquated splat, I don't see value in keeping it.

The argument to remove vector.splat because vector.broadcast can handle either a scalar operand type or a vector operand type makes no sense to me. In fact, the scalar operand type on vector.broadcast is the one to kill I feel. Again, it's very likely that whoever added that support on vector.broadcast overlooked splat: which is also why the latter is less used. Repasting from comment above:

in fact, it's awkward that vector.broadcast right now broadcasts either a scalar or an n-d vector -- two very different types. Ideally, vector.broadcast should only support *vector typed* operands while splat supports scalar operands. The fact that the latter scenario is often the common case with direct hardware support (or close to that) warrants retaining such an operation -- this is aligned with the purposes of vector types and vector dialect as opposed to tensors. (Not that it's important now, in fact, when I introduced the splat op, the intent was to mainly use it only for vector types.)

[mehdi_amini]

Is it difficult to match a vector.broadcast that is a splat? Is there code that would need to be matching both splat and broadcast when it could handle just broadcast generically?
To me these are the important questions when considering the value of having two operations instead of one.
(I don't know the answer for vector.splat vs vector.broadcast)

mlir/include/mlir/Dialect/StandardOps/IR/Ops.td

def SelectOp : Std_Op<"select", [NoSideEffect,
   let results = (outs AnyType:$result);
 
   let hasCanonicalizer = 1;
   let hasFolder = 1;
   let hasVerifier = 1;
 }
 
 //===----------------------------------------------------------------------===//
-// SplatOp
-//===----------------------------------------------------------------------===//
-
-def SplatOp : Std_Op<"splat", [NoSideEffect,
-     TypesMatchWith<"operand type matches element type of result",
-                    "aggregate", "input",
-                    "$_self.cast<ShapedType>().getElementType()">]> {
-  let summary = "splat or broadcast operation";
-  let description = [{
-    Broadcast the operand to all elements of the result vector or tensor. The
-    operand has to be of integer/index/float type. When the result is a tensor,
-    it has to be statically shaped.
-
-    Example:
-
-    ```mlir
-    %s = load %A[%i] : memref<128xf32>
-    %v = splat %s : vector<4xf32>
-    %t = splat %s : tensor<8x16xi32>
-    ```
-
-    TODO: This operation is easy to extend to broadcast to dynamically shaped
-    tensors in the same way dynamically shaped memrefs are handled.
-
-    ```mlir
-    // Broadcasts %s to a 2-d dynamically shaped tensor, with %m, %n binding
-    // to the sizes of the two dynamic dimensions.
-    %m = "foo"() : () -> (index)
-    %n = "bar"() : () -> (index)
-    %t = splat %s [%m, %n] : tensor<?x?xi32>
-    ```
-  }];
-
-  let arguments = (ins AnyTypeOf<[AnySignlessInteger, Index, AnyFloat],
-                                 "integer/index/float type">:$input);
-  let results = (outs AnyTypeOf<[AnyVectorOfAnyRank,
-                                 AnyStaticShapeTensor]>:$aggregate);
-
-  let builders = [
-    OpBuilder<(ins "Value":$element, "Type":$aggregateType),
-    [{ build($_builder, $_state, aggregateType, element); }]>];
-
-  let hasFolder = 1;
-  let hasVerifier = 1;
-
-  let assemblyFormat = "$input attr-dict `:` type($aggregate)";
-}
-
-//===----------------------------------------------------------------------===//
 // SwitchOp
 //===----------------------------------------------------------------------===//
 
 def SwitchOp : Std_Op<"switch",
     [AttrSizedOperandSegments,
      DeclareOpInterfaceMethods<BranchOpInterface, ["getSuccessorForOperands"]>,
      NoSideEffect, Terminator]> {
   let summary = "switch operation";

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

OpBuilder<(ins "Type":$resultType, "Value":$source,
       CArg<"ArrayRef<NamedAttribute>", "{}">:$attrs)>,
   ];
 
   let hasCanonicalizer = 1;
   let hasFolder = 1;
   let hasVerifier = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// SplatOp
+//===----------------------------------------------------------------------===//
+
+def Tensor_SplatOp : Tensor_Op<"splat", [
+    NoSideEffect,
+    TypesMatchWith<"operand type matches element type of result",
+                   "aggregate", "input",
+                   "$_self.cast<TensorType>().getElementType()">
+  ]> {
+  let summary = "tensor splat or broadcast operation";
+  let description = [{
+    Broadcast the operand to all elements of the result tensor. The operand is
+    required to be of integer/index/float type, and the result tensor must be
+    statically shaped.
+
+    Example:
+
+    ```mlir
+    %s = arith.constant 10.1 : f32
+    %t = tensor.splat %s : tensor<8x16xi32>
+    ```
+
+    TODO: This operation is easy to extend to broadcast to dynamically shaped
+          tensors:
+
+    ```mlir
+    // Broadcasts %s to a 2-d dynamically shaped tensor, with %m, %n binding
+    // to the sizes of the two dynamic dimensions.
+    %m = "foo"() : () -> (index)
+    %n = "bar"() : () -> (index)
+    %t = tensor.splat %s [%m, %n] : tensor<?x?xi32>
+    ```
+  }];
+
+  let arguments = (ins AnyTypeOf<[AnySignlessInteger, Index, AnyFloat],
+                                 "integer/index/float type">:$input);
+  let results = (outs AnyStaticShapeTensor:$aggregate);
+
+  let builders = [
+    OpBuilder<(ins "Value":$element, "Type":$aggregateType),
+    [{ build($_builder, $_state, aggregateType, element); }]>];
+  let assemblyFormat = "$input attr-dict `:` type($aggregate)";
+
+  let hasFolder = 1;
+}
 
 //===----------------------------------------------------------------------===//
 // YieldOp
 //===----------------------------------------------------------------------===//
 
 def Tensor_YieldOp : Tensor_Op<"yield",
     [NoSideEffect, ReturnLike, Terminator,
      HasParent<"::mlir::tensor::GenerateOp, ::mlir::tensor::PadOp">]> {

mlir/include/mlir/Dialect/Vector/IR/VectorOps.td

let description = [{
     %1 = vector.flat_transpose %0 { rows = 4: i32, columns = 4: i32 }
        : (vector<16xf32>) -> vector<16xf32>
     ```
   }];
   let assemblyFormat = "$matrix attr-dict `:` type($matrix) `->` type($res)";
 }
 
 //===----------------------------------------------------------------------===//
+// SplatOp
+//===----------------------------------------------------------------------===//
+
+def Vector_SplatOp : Vector_Op<"splat", [
+    NoSideEffect,
+    TypesMatchWith<"operand type matches element type of result",
+                   "aggregate", "input",
+                   "$_self.cast<VectorType>().getElementType()">
+  ]> {
+  let summary = "vector splat or broadcast operation";
+  let description = [{
+    Broadcast the operand to all elements of the result vector. The operand is
+    required to be of integer/index/float type.
+
+    Example:
+
+    ```mlir
+    %s = arith.constant 10.1 : f32
+    %t = vector.splat %s : vector<8x16xi32>
+    ```
+  }];
+
+  let arguments = (ins AnyTypeOf<[AnySignlessInteger, Index, AnyFloat],
+                                 "integer/index/float type">:$input);
+  let results = (outs AnyVectorOfAnyRank:$aggregate);
+
+  let builders = [
+    OpBuilder<(ins "Value":$element, "Type":$aggregateType),
+    [{ build($_builder, $_state, aggregateType, element); }]>];
+  let assemblyFormat = "$input attr-dict `:` type($aggregate)";
+
+  let hasFolder = 1;
+}
+
+//===----------------------------------------------------------------------===//
 // VectorScaleOp
 //===----------------------------------------------------------------------===//
 
 // TODO: In the future, we might want to have scalable vectors with different
 //       scales for different dimensions. E.g.: vector<[16]x[16]xf32>, in
 //       which case we might need to add an index to 'vscale' to select one
 //       of them. In order to support GPUs, we might also want to differentiate
 //       between a 'global' scale, a scale that's fixed throughout the

mlir/include/mlir/IR/Attributes.h

public:
   bool operator!=(Attribute other) const { return !(*this == other); }
   explicit operator bool() const { return impl; }
 
   bool operator!() const { return impl == nullptr; }
 
   template <typename U> bool isa() const;
   template <typename First, typename Second, typename... Rest>
   bool isa() const;
+  template <typename First, typename... Rest>
+  bool isa_and_nonnull() const;
   template <typename U> U dyn_cast() const;
   template <typename U> U dyn_cast_or_null() const;
   template <typename U> U cast() const;
 
   // Support dyn_cast'ing Attribute to itself.
   static bool classof(Attribute) { return true; }
 
   /// Return a unique identifier for the concrete attribute type. This is used
template <typename U> bool Attribute::isa() const {
   return U::classof(*this);
 }
 
 template <typename First, typename Second, typename... Rest>
 bool Attribute::isa() const {
   return isa<First>() || isa<Second, Rest...>();
 }
 
+template <typename First, typename... Rest>
+bool Attribute::isa_and_nonnull() const {
+  return impl && isa<First, Rest...>();
+}
+
 template <typename U> U Attribute::dyn_cast() const {
   return isa<U>() ? U(impl) : U(nullptr);
 }
 template <typename U> U Attribute::dyn_cast_or_null() const {
   return (impl && isa<U>()) ? U(impl) : U(nullptr);
 }
 template <typename U> U Attribute::cast() const {
   assert(isa<U>());

mlir/lib/Conversion/StandardToLLVM/StandardToLLVM.cpp

struct CondBranchOpLowering
     : public OneToOneLLVMTerminatorLowering<CondBranchOp, LLVM::CondBrOp> {
   using Super::Super;
 };
 struct SwitchOpLowering
     : public OneToOneLLVMTerminatorLowering<SwitchOp, LLVM::SwitchOp> {
   using Super::Super;
 };
 
-// The Splat operation is lowered to an insertelement + a shufflevector
-// operation. Splat to only 0-d and 1-d vector result types are lowered.
-struct SplatOpLowering : public ConvertOpToLLVMPattern<SplatOp> {
-  using ConvertOpToLLVMPattern<SplatOp>::ConvertOpToLLVMPattern;
-
-  LogicalResult
-  matchAndRewrite(SplatOp splatOp, OpAdaptor adaptor,
-                  ConversionPatternRewriter &rewriter) const override {
-    VectorType resultType = splatOp.getType().dyn_cast<VectorType>();
-    if (!resultType || resultType.getRank() > 1)
-      return failure();
-
-    // First insert it into an undef vector so we can shuffle it.
-    auto vectorType = typeConverter->convertType(splatOp.getType());
-    Value undef = rewriter.create<LLVM::UndefOp>(splatOp.getLoc(), vectorType);
-    auto zero = rewriter.create<LLVM::ConstantOp>(
-        splatOp.getLoc(),
-        typeConverter->convertType(rewriter.getIntegerType(32)),
-        rewriter.getZeroAttr(rewriter.getIntegerType(32)));
-
-    // For 0-d vector, we simply do `insertelement`.
-    if (resultType.getRank() == 0) {
-      rewriter.replaceOpWithNewOp<LLVM::InsertElementOp>(
-          splatOp, vectorType, undef, adaptor.getInput(), zero);
-      return success();
-    }
-
-    // For 1-d vector, we additionally do a `vectorshuffle`.
-    auto v = rewriter.create<LLVM::InsertElementOp>(
-        splatOp.getLoc(), vectorType, undef, adaptor.getInput(), zero);
-
-    int64_t width = splatOp.getType().cast<VectorType>().getDimSize(0);
-    SmallVector<int32_t, 4> zeroValues(width, 0);
-
-    // Shuffle the value across the desired number of elements.
-    ArrayAttr zeroAttrs = rewriter.getI32ArrayAttr(zeroValues);
-    rewriter.replaceOpWithNewOp<LLVM::ShuffleVectorOp>(splatOp, v, undef,
-                                                       zeroAttrs);
-    return success();
-  }
-};
-
-// The Splat operation is lowered to an insertelement + a shufflevector
-// operation. Splat to only 2+-d vector result types are lowered by the
-// SplatNdOpLowering, the 1-d case is handled by SplatOpLowering.
-struct SplatNdOpLowering : public ConvertOpToLLVMPattern<SplatOp> {
-  using ConvertOpToLLVMPattern<SplatOp>::ConvertOpToLLVMPattern;
-
-  LogicalResult
-  matchAndRewrite(SplatOp splatOp, OpAdaptor adaptor,
-                  ConversionPatternRewriter &rewriter) const override {
-    VectorType resultType = splatOp.getType().dyn_cast<VectorType>();
-    if (!resultType || resultType.getRank() <= 1)
-      return failure();
-
-    // First insert it into an undef vector so we can shuffle it.
-    auto loc = splatOp.getLoc();
-    auto vectorTypeInfo =
-        LLVM::detail::extractNDVectorTypeInfo(resultType, *getTypeConverter());
-    auto llvmNDVectorTy = vectorTypeInfo.llvmNDVectorTy;
-    auto llvm1DVectorTy = vectorTypeInfo.llvm1DVectorTy;
-    if (!llvmNDVectorTy || !llvm1DVectorTy)
-      return failure();
-
-    // Construct returned value.
-    Value desc = rewriter.create<LLVM::UndefOp>(loc, llvmNDVectorTy);
-
-    // Construct a 1-D vector with the splatted value that we insert in all the
-    // places within the returned descriptor.
-    Value vdesc = rewriter.create<LLVM::UndefOp>(loc, llvm1DVectorTy);
-    auto zero = rewriter.create<LLVM::ConstantOp>(
-        loc, typeConverter->convertType(rewriter.getIntegerType(32)),
-        rewriter.getZeroAttr(rewriter.getIntegerType(32)));
-    Value v = rewriter.create<LLVM::InsertElementOp>(loc, llvm1DVectorTy, vdesc,
-                                                     adaptor.getInput(), zero);
-
-    // Shuffle the value across the desired number of elements.
-    int64_t width = resultType.getDimSize(resultType.getRank() - 1);
-    SmallVector<int32_t, 4> zeroValues(width, 0);
-    ArrayAttr zeroAttrs = rewriter.getI32ArrayAttr(zeroValues);
-    v = rewriter.create<LLVM::ShuffleVectorOp>(loc, v, v, zeroAttrs);
-
-    // Iterate of linear index, convert to coords space and insert splatted 1-D
-    // vector in each position.
-    nDVectorIterate(vectorTypeInfo, rewriter, [&](ArrayAttr position) {
-      desc = rewriter.create<LLVM::InsertValueOp>(loc, llvmNDVectorTy, desc, v,
-                                                  position);
-    });
-    rewriter.replaceOp(splatOp, desc);
-    return success();
-  }
-};
-
 } // namespace
 
 void mlir::populateStdToLLVMFuncOpConversionPattern(
     LLVMTypeConverter &converter, RewritePatternSet &patterns) {
   if (converter.getOptions().useBarePtrCallConv)
     patterns.add<BarePtrFuncOpConversion>(converter);
   else
     patterns.add<FuncOpConversion>(converter);
 }
 
 void mlir::populateStdToLLVMConversionPatterns(LLVMTypeConverter &converter,
                                                RewritePatternSet &patterns) {
   populateStdToLLVMFuncOpConversionPattern(converter, patterns);
   // clang-format off
   patterns.add<
       AssertOpLowering,
       BranchOpLowering,
       CallIndirectOpLowering,
       CallOpLowering,
       CondBranchOpLowering,
       ConstantOpLowering,
       ReturnOpLowering,
       SelectOpLowering,
-      SplatOpLowering,
-      SplatNdOpLowering,
       SwitchOpLowering>(converter);
   // clang-format on
 }
 
 namespace {
 /// A pass converting MLIR operations into the LLVM IR dialect.
 struct LLVMLoweringPass : public ConvertStandardToLLVMBase<LLVMLoweringPass> {
   LLVMLoweringPass() = default;

mlir/lib/Conversion/StandardToSPIRV/StandardToSPIRV.cpp

 class SelectOpPattern final : public OpConversionPattern<SelectOp> {
 public:
   using OpConversionPattern<SelectOp>::OpConversionPattern;
   LogicalResult
   matchAndRewrite(SelectOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override;
 };
 
-/// Converts std.splat to spv.CompositeConstruct.
-class SplatPattern final : public OpConversionPattern<SplatOp> {
-public:
-  using OpConversionPattern<SplatOp>::OpConversionPattern;
-
-  LogicalResult
-  matchAndRewrite(SplatOp op, OpAdaptor adaptor,
-                  ConversionPatternRewriter &rewriter) const override;
-};
-
 /// Converts std.br to spv.Branch.
 struct BranchOpPattern final : public OpConversionPattern<BranchOp> {
   using OpConversionPattern<BranchOp>::OpConversionPattern;
 
   LogicalResult
   matchAndRewrite(BranchOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override;
 };
SelectOpPattern::matchAndRewrite(SelectOp op, OpAdaptor adaptor,
                                  ConversionPatternRewriter &rewriter) const {
   rewriter.replaceOpWithNewOp<spirv::SelectOp>(op, adaptor.getCondition(),
                                                adaptor.getTrueValue(),
                                                adaptor.getFalseValue());
   return success();
 }
 
 //===----------------------------------------------------------------------===//
-// SplatOp
-//===----------------------------------------------------------------------===//
-
-LogicalResult
-SplatPattern::matchAndRewrite(SplatOp op, OpAdaptor adaptor,
-                              ConversionPatternRewriter &rewriter) const {
-  auto dstVecType = op.getType().dyn_cast<VectorType>();
-  if (!dstVecType || !spirv::CompositeType::isValid(dstVecType))
-    return failure();
-  SmallVector<Value, 4> source(dstVecType.getNumElements(), adaptor.getInput());
-  rewriter.replaceOpWithNewOp<spirv::CompositeConstructOp>(op, dstVecType,
-                                                           source);
-  return success();
-}
-
-//===----------------------------------------------------------------------===//
 // BranchOpPattern
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 BranchOpPattern::matchAndRewrite(BranchOp op, OpAdaptor adaptor,
                                  ConversionPatternRewriter &rewriter) const {
   rewriter.replaceOpWithNewOp<spirv::BranchOp>(op, op.getDest(),
                                                adaptor.getDestOperands());
patterns.add<
       // Unary and binary patterns
       spirv::ElementwiseOpPattern<arith::MaxFOp, spirv::GLSLFMaxOp>,
       spirv::ElementwiseOpPattern<arith::MaxSIOp, spirv::GLSLSMaxOp>,
       spirv::ElementwiseOpPattern<arith::MaxUIOp, spirv::GLSLUMaxOp>,
       spirv::ElementwiseOpPattern<arith::MinFOp, spirv::GLSLFMinOp>,
       spirv::ElementwiseOpPattern<arith::MinSIOp, spirv::GLSLSMinOp>,
       spirv::ElementwiseOpPattern<arith::MinUIOp, spirv::GLSLUMinOp>,
 
-      ReturnOpPattern, SelectOpPattern, SplatPattern, BranchOpPattern,
-      CondBranchOpPattern>(typeConverter, context);
+      ReturnOpPattern, SelectOpPattern, BranchOpPattern, CondBranchOpPattern>(
+      typeConverter, context);
 }
 
 void populateTensorToSPIRVPatterns(SPIRVTypeConverter &typeConverter,
                                    int64_t byteCountThreshold,
                                    RewritePatternSet &patterns) {
   patterns.add<TensorExtractPattern>(typeConverter, patterns.getContext(),
                                      byteCountThreshold);
 }
 
 } // namespace mlir

mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp

LogicalResult matchAndRewrite(FMAOp op,
     auto vType = op.getVectorType();
     if (vType.getRank() < 2)
       return failure();
 
     auto loc = op.getLoc();
     auto elemType = vType.getElementType();
     Value zero = rewriter.create<arith::ConstantOp>(
         loc, elemType, rewriter.getZeroAttr(elemType));
-    Value desc = rewriter.create<SplatOp>(loc, vType, zero);
+    Value desc = rewriter.create<vector::SplatOp>(loc, vType, zero);
     for (int64_t i = 0, e = vType.getShape().front(); i != e; ++i) {
       Value extrLHS = rewriter.create<ExtractOp>(loc, op.lhs(), i);
       Value extrRHS = rewriter.create<ExtractOp>(loc, op.rhs(), i);
       Value extrACC = rewriter.create<ExtractOp>(loc, op.acc(), i);
       Value fma = rewriter.create<FMAOp>(loc, extrLHS, extrRHS, extrACC);
       desc = rewriter.create<InsertOp>(loc, fma, desc, i);
     }
     rewriter.replaceOp(op, desc);
private:
   // Helper to emit a call.
   static void emitCall(ConversionPatternRewriter &rewriter, Location loc,
                        Operation *ref, ValueRange params = ValueRange()) {
     rewriter.create<LLVM::CallOp>(loc, TypeRange(), SymbolRefAttr::get(ref),
                                   params);
   }
 };
 
+/// The Splat operation is lowered to an insertelement + a shufflevector
+/// operation. Splat to only 0-d and 1-d vector result types are lowered.
+struct VectorSplatOpLowering : public ConvertOpToLLVMPattern<vector::SplatOp> {
+  using ConvertOpToLLVMPattern<vector::SplatOp>::ConvertOpToLLVMPattern;
+
+  LogicalResult
+  matchAndRewrite(vector::SplatOp splatOp, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    VectorType resultType = splatOp.getType().cast<VectorType>();
+    if (resultType.getRank() > 1)
+      return failure();
+
+    // First insert it into an undef vector so we can shuffle it.
+    auto vectorType = typeConverter->convertType(splatOp.getType());
+    Value undef = rewriter.create<LLVM::UndefOp>(splatOp.getLoc(), vectorType);
+    auto zero = rewriter.create<LLVM::ConstantOp>(
+        splatOp.getLoc(),
+        typeConverter->convertType(rewriter.getIntegerType(32)),
+        rewriter.getZeroAttr(rewriter.getIntegerType(32)));
+
+    // For 0-d vector, we simply do `insertelement`.
+    if (resultType.getRank() == 0) {
+      rewriter.replaceOpWithNewOp<LLVM::InsertElementOp>(
+          splatOp, vectorType, undef, adaptor.input(), zero);
+      return success();
+    }
+
+    // For 1-d vector, we additionally do a `vectorshuffle`.
+    auto v = rewriter.create<LLVM::InsertElementOp>(
+        splatOp.getLoc(), vectorType, undef, adaptor.input(), zero);
+
+    int64_t width = splatOp.getType().cast<VectorType>().getDimSize(0);
+    SmallVector<int32_t, 4> zeroValues(width, 0);
+
+    // Shuffle the value across the desired number of elements.
+    ArrayAttr zeroAttrs = rewriter.getI32ArrayAttr(zeroValues);
+    rewriter.replaceOpWithNewOp<LLVM::ShuffleVectorOp>(splatOp, v, undef,
+                                                       zeroAttrs);
+    return success();
+  }
+};
+
+/// The Splat operation is lowered to an insertelement + a shufflevector
+/// operation. Splat to only 2+-d vector result types are lowered by the
+/// SplatNdOpLowering, the 1-d case is handled by SplatOpLowering.
+struct VectorSplatNdOpLowering : public ConvertOpToLLVMPattern<SplatOp> {
+  using ConvertOpToLLVMPattern<SplatOp>::ConvertOpToLLVMPattern;
+
+  LogicalResult
+  matchAndRewrite(SplatOp splatOp, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    VectorType resultType = splatOp.getType();
+    if (resultType.getRank() <= 1)
+      return failure();
+
+    // First insert it into an undef vector so we can shuffle it.
+    auto loc = splatOp.getLoc();
+    auto vectorTypeInfo =
+        LLVM::detail::extractNDVectorTypeInfo(resultType, *getTypeConverter());
+    auto llvmNDVectorTy = vectorTypeInfo.llvmNDVectorTy;
+    auto llvm1DVectorTy = vectorTypeInfo.llvm1DVectorTy;
+    if (!llvmNDVectorTy || !llvm1DVectorTy)
+      return failure();
+
+    // Construct returned value.
+    Value desc = rewriter.create<LLVM::UndefOp>(loc, llvmNDVectorTy);
+
+    // Construct a 1-D vector with the splatted value that we insert in all the
+    // places within the returned descriptor.
+    Value vdesc = rewriter.create<LLVM::UndefOp>(loc, llvm1DVectorTy);
+    auto zero = rewriter.create<LLVM::ConstantOp>(
+        loc, typeConverter->convertType(rewriter.getIntegerType(32)),
+        rewriter.getZeroAttr(rewriter.getIntegerType(32)));
+    Value v = rewriter.create<LLVM::InsertElementOp>(loc, llvm1DVectorTy, vdesc,
+                                                     adaptor.input(), zero);
+
+    // Shuffle the value across the desired number of elements.
+    int64_t width = resultType.getDimSize(resultType.getRank() - 1);
+    SmallVector<int32_t, 4> zeroValues(width, 0);
+    ArrayAttr zeroAttrs = rewriter.getI32ArrayAttr(zeroValues);
+    v = rewriter.create<LLVM::ShuffleVectorOp>(loc, v, v, zeroAttrs);
+
+    // Iterate of linear index, convert to coords space and insert splatted 1-D
+    // vector in each position.
+    nDVectorIterate(vectorTypeInfo, rewriter, [&](ArrayAttr position) {
+      desc = rewriter.create<LLVM::InsertValueOp>(loc, llvmNDVectorTy, desc, v,
+                                                  position);
+    });
+    rewriter.replaceOp(splatOp, desc);
+    return success();
+  }
+};
+
 } // namespace
 
 /// Populate the given list with patterns that convert from Vector to LLVM.
 void mlir::populateVectorToLLVMConversionPatterns(
     LLVMTypeConverter &converter, RewritePatternSet &patterns,
     bool reassociateFPReductions) {
   MLIRContext *ctx = converter.getDialect()->getContext();
   patterns.add<VectorFMAOpNDRewritePattern>(ctx);
   populateVectorInsertExtractStridedSliceTransforms(patterns);
   patterns.add<VectorReductionOpConversion>(converter, reassociateFPReductions);
   patterns
       .add<VectorBitCastOpConversion, VectorShuffleOpConversion,
            VectorExtractElementOpConversion, VectorExtractOpConversion,
            VectorFMAOp1DConversion, VectorInsertElementOpConversion,
            VectorInsertOpConversion, VectorPrintOpConversion,
            VectorTypeCastOpConversion, VectorScaleOpConversion,
            VectorLoadStoreConversion<vector::LoadOp, vector::LoadOpAdaptor>,
            VectorLoadStoreConversion<vector::MaskedLoadOp,
                                      vector::MaskedLoadOpAdaptor>,
            VectorLoadStoreConversion<vector::StoreOp, vector::StoreOpAdaptor>,
            VectorLoadStoreConversion<vector::MaskedStoreOp,
                                      vector::MaskedStoreOpAdaptor>,
            VectorGatherOpConversion, VectorScatterOpConversion,
-           VectorExpandLoadOpConversion, VectorCompressStoreOpConversion>(
-          converter);
+           VectorExpandLoadOpConversion, VectorCompressStoreOpConversion,
+           VectorSplatOpLowering, VectorSplatNdOpLowering>(converter);
   // Transfer ops with rank > 1 are handled by VectorToSCF.
   populateVectorTransferLoweringPatterns(patterns, /*maxTransferRank=*/1);
 }
 
 void mlir::populateVectorToLLVMMatrixConversionPatterns(
     LLVMTypeConverter &converter, RewritePatternSet &patterns) {
   patterns.add<VectorMatmulOpConversion>(converter);
   patterns.add<VectorFlatTransposeOpConversion>(converter);
 }

mlir/lib/Conversion/VectorToSCF/VectorToSCF.cpp

static Value handleOutOfBoundsDim(OpBuilder &b, TransferReadOp xferOp,
                                     ValueRange /*loopState*/) {
     SmallVector<Value, 8> storeIndices;
     getBufferIndices(xferOp, storeIndices);
     storeIndices.push_back(iv);
 
     Location loc = xferOp.getLoc();
     auto bufferType = buffer.getType().dyn_cast<ShapedType>();
     auto vecType = bufferType.getElementType().dyn_cast<VectorType>();
-    auto vec = b.create<SplatOp>(loc, vecType, xferOp.padding());
+    auto vec = b.create<vector::SplatOp>(loc, vecType, xferOp.padding());
     b.create<memref::StoreOp>(loc, vec, buffer, storeIndices);
 
     return Value();
   }
 
   /// Cleanup after rewriting the op.
   static void cleanup(PatternRewriter &rewriter, TransferReadOp xferOp,
                       scf::ForOp /*forOp*/) {
struct UnrollTransferReadConversion
 
   /// Return the vector into which the newly created TransferReadOp results
   /// are inserted.
   Value getResultVector(TransferReadOp xferOp,
                         PatternRewriter &rewriter) const {
     if (auto insertOp = getInsertOp(xferOp))
       return insertOp.dest();
     Location loc = xferOp.getLoc();
-    return rewriter.create<SplatOp>(loc, xferOp.getVectorType(),
+    return rewriter.create<vector::SplatOp>(loc, xferOp.getVectorType(),
                                             xferOp.padding());
   }
 
   /// If the result of the TransferReadOp has exactly one user, which is a
   /// vector::InsertOp, return that operation.
   vector::InsertOp getInsertOp(TransferReadOp xferOp) const {
     if (xferOp->hasOneUse()) {
       Operation *xferOpUser = *xferOp->getUsers().begin();
       if (auto insertOp = dyn_cast<vector::InsertOp>(xferOpUser))
auto nextVec = generateInBoundsCheck(
         /*outOfBoundsCase=*/
         [&](OpBuilder & /*b*/, Location loc) { return vec; });
     b.create<scf::YieldOp>(loc, nextVec);
   }
 
   static Value initialLoopState(OpBuilder &b, TransferReadOp xferOp) {
     // Inititalize vector with padding value.
     Location loc = xferOp.getLoc();
-    return b.create<SplatOp>(loc, xferOp.getVectorType(), xferOp.padding());
+    return b.create<vector::SplatOp>(loc, xferOp.getVectorType(),
+                                     xferOp.padding());
   }
 };
 
 /// Codegen strategy for TransferWriteOp.
 template <>
 struct Strategy1d<TransferWriteOp> {
   static void generateForLoopBody(OpBuilder &b, Location loc,
                                   TransferWriteOp xferOp, Value iv,

mlir/lib/Conversion/VectorToSPIRV/VectorToSPIRV.cpp

matchAndRewrite(vector::InsertStridedSliceOp insertOp, OpAdaptor adaptor,
     rewriter.replaceOpWithNewOp<spirv::VectorShuffleOp>(
         insertOp, dstVector.getType(), dstVector, srcVector,
         rewriter.getI32ArrayAttr(indices));
 
     return success();
   }
 };
 
+class VectorSplatPattern final : public OpConversionPattern<vector::SplatOp> {
+public:
+  using OpConversionPattern<vector::SplatOp>::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(vector::SplatOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    VectorType dstVecType = op.getType();
+    if (!spirv::CompositeType::isValid(dstVecType))
+      return failure();
+    SmallVector<Value, 4> source(dstVecType.getNumElements(), adaptor.input());
+    rewriter.replaceOpWithNewOp<spirv::CompositeConstructOp>(op, dstVecType,
+                                                             source);
+    return success();
+  }
+};
+
 } // namespace
 
 void mlir::populateVectorToSPIRVPatterns(SPIRVTypeConverter &typeConverter,
                                          RewritePatternSet &patterns) {
   patterns.add<VectorBitcastConvert, VectorBroadcastConvert,
                VectorExtractElementOpConvert, VectorExtractOpConvert,
                VectorExtractStridedSliceOpConvert, VectorFmaOpConvert,
                VectorInsertElementOpConvert, VectorInsertOpConvert,
-               VectorInsertStridedSliceOpConvert>(typeConverter,
-                                                  patterns.getContext());
+               VectorInsertStridedSliceOpConvert, VectorSplatPattern>(
+      typeConverter, patterns.getContext());
 }

mlir/lib/Dialect/StandardOps/IR/Ops.cpp

if (conditionType != shapedConditionType)
     return emitOpError() << "expected condition type to have the same shape "
                             "as the result type, expected "
                          << shapedConditionType << ", but got "
                          << conditionType;
   return success();
 }
 
 //===----------------------------------------------------------------------===//
-// SplatOp
-//===----------------------------------------------------------------------===//
-
-LogicalResult SplatOp::verify() {
-  // TODO: we could replace this by a trait.
-  if (getOperand().getType() != getType().cast<ShapedType>().getElementType())
-    return emitError("operand should be of elemental type of result type");
-
-  return success();
-}
-
-// Constant folding hook for SplatOp.
-OpFoldResult SplatOp::fold(ArrayRef<Attribute> operands) {
-  assert(operands.size() == 1 && "splat takes one operand");
-
-  auto constOperand = operands.front();
-  if (!constOperand || !constOperand.isa<IntegerAttr, FloatAttr>())
-    return {};
-
-  auto shapedType = getType().cast<ShapedType>();
-  assert(shapedType.getElementType() == constOperand.getType() &&
-         "incorrect input attribute type for folding");
-
-  // SplatElementsAttr::get treats single value for second arg as being a splat.
-  return SplatElementsAttr::get(shapedType, {constOperand});
-}
-
-//===----------------------------------------------------------------------===//
 // SwitchOp
 //===----------------------------------------------------------------------===//
 
 void SwitchOp::build(OpBuilder &builder, OperationState &result, Value value,
                      Block *defaultDestination, ValueRange defaultOperands,
                      DenseIntElementsAttr caseValues,
                      BlockRange caseDestinations,
                      ArrayRef<ValueRange> caseOperands) {

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

 OpFoldResult PadOp::fold(ArrayRef<Attribute>) {
   if (getResultType().hasStaticShape() && getResultType() == getSourceType() &&
       !nofold())
     return source();
   return {};
 }
 
 //===----------------------------------------------------------------------===//
+// SplatOp
+//===----------------------------------------------------------------------===//
+
+OpFoldResult SplatOp::fold(ArrayRef<Attribute> operands) {
+  auto constOperand = operands.front();
+  if (!constOperand.isa_and_nonnull<IntegerAttr, FloatAttr>())
+    return {};
+
+  // SplatElementsAttr::get treats single value for second arg as being a splat.
+  return SplatElementsAttr::get(getType(), {constOperand});
+}
+
+//===----------------------------------------------------------------------===//
 // TableGen'd op method definitions
 //===----------------------------------------------------------------------===//
 
 #define GET_OP_CLASSES
 #include "mlir/Dialect/Tensor/IR/TensorOps.cpp.inc"

mlir/lib/Dialect/Vector/IR/VectorOps.cpp

 // operations, in particular super-vector loads and stores.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Vector/IR/VectorOps.h"
 
 #include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/StandardOps/Utils/Utils.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Utils/IndexingUtils.h"
 #include "mlir/Dialect/Utils/StructuredOpsUtils.h"
 #include "mlir/IR/AffineExpr.h"
 #include "mlir/IR/AffineMap.h"
 #include "mlir/IR/BlockAndValueMapping.h"
 #include "mlir/IR/Builders.h"

 public:
   using OpRewritePattern<ExtractStridedSliceOp>::OpRewritePattern;
 
   LogicalResult matchAndRewrite(ExtractStridedSliceOp op,
                                 PatternRewriter &rewriter) const override {
     auto splat = op.vector().getDefiningOp<SplatOp>();
     if (!splat)
       return failure();
-    rewriter.replaceOpWithNewOp<SplatOp>(op, op.getType(), splat.getInput());
+    rewriter.replaceOpWithNewOp<SplatOp>(op, op.getType(), splat.input());
     return success();
   }
 };
 
 } // namespace
 
 void ExtractStridedSliceOp::getCanonicalizationPatterns(
     RewritePatternSet &results, MLIRContext *context) {
void mlir::vector::populateVectorToVectorCanonicalizationPatterns(
     RewritePatternSet &patterns) {
   patterns
       .add<CreateMaskFolder, MaskedLoadFolder, MaskedStoreFolder, GatherFolder,
            ScatterFolder, ExpandLoadFolder, CompressStoreFolder,
            StridedSliceConstantMaskFolder, TransposeFolder>(
           patterns.getContext());
 }
 
+//===----------------------------------------------------------------------===//
+// SplatOp
+//===----------------------------------------------------------------------===//
+
+OpFoldResult SplatOp::fold(ArrayRef<Attribute> operands) {
+  auto constOperand = operands.front();
+  if (!constOperand.isa_and_nonnull<IntegerAttr, FloatAttr>())
+    return {};
+
+  // SplatElementsAttr::get treats single value for second arg as being a splat.
+  return SplatElementsAttr::get(getType(), {constOperand});
+}
+
+//===----------------------------------------------------------------------===//
+// TableGen'd op method definitions
+//===----------------------------------------------------------------------===//
+
 #define GET_OP_CLASSES
 #include "mlir/Dialect/Vector/IR/VectorOps.cpp.inc"

mlir/lib/Dialect/Vector/Transforms/VectorInsertExtractStridedSliceRewritePatterns.cpp

 //===- VectorInsertExtractStridedSliceRewritePatterns.cpp - Rewrites ------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Utils/IndexingUtils.h"
 #include "mlir/Dialect/Vector/IR/VectorOps.h"
 #include "mlir/Dialect/Vector/Transforms/VectorRewritePatterns.h"
 #include "mlir/Dialect/Vector/Utils/VectorUtils.h"
 #include "mlir/IR/BuiltinTypes.h"
 
 using namespace mlir;
 using namespace mlir::vector;

mlir/lib/Dialect/Vector/Transforms/VectorTransferOpTransforms.cpp

 //===- VectorTransferOpTransforms.cpp - transfer op transforms ------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements functions concerned with optimizing transfer_read and
 // transfer_write ops.
 //
 //===----------------------------------------------------------------------===//
+
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Vector/IR/VectorOps.h"
 #include "mlir/Dialect/Vector/Transforms/VectorTransforms.h"
 #include "mlir/Dialect/Vector/Utils/VectorUtils.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/Dominance.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Support/Debug.h"

mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp

 
 #include <type_traits>
 
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/SCF/SCF.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Utils/StructuredOpsUtils.h"
 
 #include "mlir/Dialect/Vector/Transforms/VectorTransforms.h"
 #include "mlir/IR/ImplicitLocOpBuilder.h"
 #include "mlir/IR/Matchers.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/Interfaces/VectorInterfaces.h"
 
LogicalResult matchAndRewrite(vector::BroadcastOp op,
                                 PatternRewriter &rewriter) const override {
     auto loc = op.getLoc();
     VectorType dstType = op.getVectorType();
     VectorType srcType = op.getSourceType().dyn_cast<VectorType>();
     Type eltType = dstType.getElementType();
 
     // Scalar to any vector can use splat.
     if (!srcType) {
-      rewriter.replaceOpWithNewOp<SplatOp>(op, dstType, op.source());
+      rewriter.replaceOpWithNewOp<vector::SplatOp>(op, dstType, op.source());
       return success();
     }
 
     // Determine rank of source and destination.
     int64_t srcRank = srcType.getRank();
     int64_t dstRank = dstType.getRank();
 
     // Stretching scalar inside vector (e.g. vector<1xf32>) can use splat.
     if (srcRank <= 1 && dstRank == 1) {
       Value ext;
       if (srcRank == 0)
         ext = rewriter.create<vector::ExtractElementOp>(loc, op.source());
       else
         ext = rewriter.create<vector::ExtractOp>(loc, op.source(), 0);
-      rewriter.replaceOpWithNewOp<SplatOp>(op, dstType, ext);
+      rewriter.replaceOpWithNewOp<vector::SplatOp>(op, dstType, ext);
       return success();
     }
 
     // Duplicate this rank.
     // For example:
     //   %x = broadcast %y  : k-D to n-D, k < n
     // becomes:
     //   %b = broadcast %y  : k-D to (n-1)-D
LogicalResult matchAndRewrite(vector::TransferReadOp read,
 
     // Out-of-bounds dims are handled by MaterializeTransferMask.
     if (read.hasOutOfBoundsDim())
       return failure();
 
     // Create vector load op.
     Operation *loadOp;
     if (read.mask()) {
-      Value fill = rewriter.create<SplatOp>(
+      Value fill = rewriter.create<vector::SplatOp>(
           read.getLoc(), unbroadcastedVectorType, read.padding());
       loadOp = rewriter.create<vector::MaskedLoadOp>(
           read.getLoc(), unbroadcastedVectorType, read.source(), read.indices(),
           read.mask(), fill);
     } else {
       loadOp = rewriter.create<vector::LoadOp>(read.getLoc(),
                                                unbroadcastedVectorType,
                                                read.source(), read.indices());
static Value buildVectorComparison(PatternRewriter &rewriter, Operation *op,
   } else {
     indicesAttr = rewriter.getI64VectorAttr(
         llvm::to_vector<4>(llvm::seq<int64_t>(0, dim)));
   }
   Value indices = rewriter.create<arith::ConstantOp>(loc, indicesAttr);
   // Add in an offset if requested.
   if (off) {
     Value o = createCastToIndexLike(rewriter, loc, idxType, *off);
-    Value ov = rewriter.create<SplatOp>(loc, indices.getType(), o);
+    Value ov = rewriter.create<vector::SplatOp>(loc, indices.getType(), o);
     indices = rewriter.create<arith::AddIOp>(loc, ov, indices);
   }
   // Construct the vector comparison.
   Value bound = createCastToIndexLike(rewriter, loc, idxType, b);
-  Value bounds = rewriter.create<SplatOp>(loc, indices.getType(), bound);
+  Value bounds =
+      rewriter.create<vector::SplatOp>(loc, indices.getType(), bound);
   return rewriter.create<arith::CmpIOp>(loc, arith::CmpIPredicate::slt, indices,
                                         bounds);
 }
 
 template <typename ConcreteOp>
 struct MaterializeTransferMask : public OpRewritePattern<ConcreteOp> {
 public:
   explicit MaterializeTransferMask(MLIRContext *context, bool enableIndexOpt)

mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp

 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Vector/Utils/VectorUtils.h"
 
 #include "mlir/Dialect/Affine/Analysis/LoopAnalysis.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Arithmetic/IR/Arithmetic.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Vector/IR/VectorOps.h"
 #include "mlir/IR/Builders.h"
 #include "mlir/IR/IntegerSet.h"
 #include "mlir/IR/Operation.h"
 #include "mlir/Support/LLVM.h"
 #include "mlir/Support/MathExtras.h"
 #include <numeric>

mlir/test/Conversion/StandardToLLVM/standard-to-llvm.mlir

 // CHECK-NEXT: ^bb2:
 ^bb2:
 // CHECK-NEXT:  llvm.br ^bb1
   br ^bb1
 }
 
 // -----
 
-// CHECK-LABEL: @splat_0d
-// CHECK-SAME: %[[ARG:.*]]: f32
-func @splat_0d(%a: f32) -> vector<f32> {
-  %v = splat %a : vector<f32>
-  return %v : vector<f32>
-}
-// CHECK-NEXT: %[[UNDEF:[0-9]+]] = llvm.mlir.undef : vector<1xf32>
-// CHECK-NEXT: %[[ZERO:[0-9]+]] = llvm.mlir.constant(0 : i32) : i32
-// CHECK-NEXT: %[[V:[0-9]+]] = llvm.insertelement %[[ARG]], %[[UNDEF]][%[[ZERO]] : i32] : vector<1xf32>
-// CHECK-NEXT: llvm.return %[[V]] : vector<1xf32>
-
-// -----
-
-// CHECK-LABEL: @splat
-// CHECK-SAME: %[[A:arg[0-9]+]]: vector<4xf32>
-// CHECK-SAME: %[[ELT:arg[0-9]+]]: f32
-func @splat(%a: vector<4xf32>, %b: f32) -> vector<4xf32> {
-  %vb = splat %b : vector<4xf32>
-  %r = arith.mulf %a, %vb : vector<4xf32>
-  return %r : vector<4xf32>
-}
-// CHECK-NEXT: %[[UNDEF:[0-9]+]] = llvm.mlir.undef : vector<4xf32>
-// CHECK-NEXT: %[[ZERO:[0-9]+]] = llvm.mlir.constant(0 : i32) : i32
-// CHECK-NEXT: %[[V:[0-9]+]] = llvm.insertelement %[[ELT]], %[[UNDEF]][%[[ZERO]] : i32] : vector<4xf32>
-// CHECK-NEXT: %[[SPLAT:[0-9]+]] = llvm.shufflevector %[[V]], %[[UNDEF]] [0 : i32, 0 : i32, 0 : i32, 0 : i32]
-// CHECK-NEXT: %[[SCALE:[0-9]+]] = llvm.fmul %[[A]], %[[SPLAT]] : vector<4xf32>
-// CHECK-NEXT: llvm.return %[[SCALE]] : vector<4xf32>
-
-// -----
-
 // CHECK-LABEL: func @ceilf(
 // CHECK-SAME: f32
 func @ceilf(%arg0 : f32) {
   // CHECK: "llvm.intr.ceil"(%arg0) : (f32) -> f32
   %0 = math.ceil %arg0 : f32
   std.return
 }
 

mlir/test/Conversion/StandardToSPIRV/std-ops-to-spirv.mlir

func @tensor_extract_constant(%a : index, %b: index, %c: index) -> i32 {
   %extract = tensor.extract %cst[%a, %b, %c] : tensor<2x2x3xi32>
   // CHECK: spv.ReturnValue %[[VAL]]
   return %extract : i32
 }
 
 // -----
 
 //===----------------------------------------------------------------------===//
-// splat
-//===----------------------------------------------------------------------===//
-
-// CHECK-LABEL: func @splat
-//  CHECK-SAME: (%[[A:.+]]: f32)
-//       CHECK:   %[[VAL:.+]] = spv.CompositeConstruct %[[A]], %[[A]], %[[A]], %[[A]] : vector<4xf32>
-//       CHECK:   spv.ReturnValue %[[VAL]]
-func @splat(%f : f32) -> vector<4xf32> {
-  %splat = splat %f : vector<4xf32>
-  return %splat : vector<4xf32>
-}
-
-// -----
-
-//===----------------------------------------------------------------------===//
 // std.br, std.cond_br
 //===----------------------------------------------------------------------===//
 
 module attributes {
   spv.target_env = #spv.target_env<#spv.vce<v1.0, [], []>, {}>
 } {
 
 // CHECK-LABEL: func @simple_loop

mlir/test/Conversion/VectorToLLVM/vector-mask-to-llvm.mlir

 // RUN: mlir-opt %s --convert-vector-to-llvm='enable-index-optimizations=1' | FileCheck %s --check-prefix=CMP32
 // RUN: mlir-opt %s --convert-vector-to-llvm='enable-index-optimizations=0' | FileCheck %s --check-prefix=CMP64
 
 // CMP32-LABEL: @genbool_var_1d(
 // CMP32-SAME: %[[ARG:.*]]: index)
 // CMP32: %[[T0:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]> : vector<11xi32>
 // CMP32: %[[T1:.*]] = arith.index_cast %[[ARG]] : index to i32
-// CMP32: %[[T2:.*]] = splat %[[T1]] : vector<11xi32>
-// CMP32: %[[T3:.*]] = arith.cmpi slt, %[[T0]], %[[T2]] : vector<11xi32>
-// CMP32: return %[[T3]] : vector<11xi1>
+// CMP32: %[[T2:.*]] = llvm.insertelement %[[T1]], %{{.*}}[%{{.*}} : i32] : vector<11xi32>
+// CMP32: %[[T3:.*]] = llvm.shufflevector %[[T2]], %{{.*}} [0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32] : vector<11xi32>, vector<11xi32>
+// CMP32: %[[T4:.*]] = arith.cmpi slt, %[[T0]], %[[T3]] : vector<11xi32>
+// CMP32: return %[[T4]] : vector<11xi1>
 
 // CMP64-LABEL: @genbool_var_1d(
 // CMP64-SAME: %[[ARG:.*]]: index)
 // CMP64: %[[T0:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]> : vector<11xi64>
 // CMP64: %[[T1:.*]] = arith.index_cast %[[ARG]] : index to i64
-// CMP64: %[[T2:.*]] = splat %[[T1]] : vector<11xi64>
-// CMP64: %[[T3:.*]] = arith.cmpi slt, %[[T0]], %[[T2]] : vector<11xi64>
-// CMP64: return %[[T3]] : vector<11xi1>
+// CMP64: %[[T2:.*]] = llvm.insertelement %[[T1]], %{{.*}}[%{{.*}} : i32] : vector<11xi64>
+// CMP64: %[[T3:.*]] = llvm.shufflevector %[[T2]], %{{.*}} [0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32, 0 : i32] : vector<11xi64>, vector<11xi64>
+// CMP64: %[[T4:.*]] = arith.cmpi slt, %[[T0]], %[[T3]] : vector<11xi64>
+// CMP64: return %[[T4]] : vector<11xi1>
 
 func @genbool_var_1d(%arg0: index) -> vector<11xi1> {
   %0 = vector.create_mask %arg0 : vector<11xi1>
   return %0 : vector<11xi1>
 }
 
 // CMP32-LABEL: @transfer_read_1d
 // CMP32: %[[C:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]> : vector<16xi32>

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

 // -----
 
 func @broadcast_vec0d_from_f32(%arg0: f32) -> vector<f32> {
   %0 = vector.broadcast %arg0 : f32 to vector<f32>
   return %0 : vector<f32>
 }
 // CHECK-LABEL: @broadcast_vec0d_from_f32
 // CHECK-SAME:  %[[A:.*]]: f32)
-// CHECK:       %[[T0:.*]] = splat %[[A]] : vector<f32>
-// CHECK:       return %[[T0]] : vector<f32>
+// CHECK:       %[[T0:.*]] = llvm.insertelement %[[A]]
+// CHECK:       %[[T1:.*]] = builtin.unrealized_conversion_cast %[[T0]] : vector<1xf32> to vector<f32>
+// CHECK:       return %[[T1]] : vector<f32>
 
 // -----
 
 func @broadcast_vec0d_from_vec0d(%arg0: vector<f32>) -> vector<f32> {
   %0 = vector.broadcast %arg0 : vector<f32> to vector<f32>
   return %0 : vector<f32>
 }
 // CHECK-LABEL: @broadcast_vec0d_from_vec0d(
 // CHECK-SAME:  %[[A:.*]]: vector<f32>)
 // CHECK:       return %[[A]] : vector<f32>
 
 // -----
 
 func @broadcast_vec1d_from_f32(%arg0: f32) -> vector<2xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2xf32>
   return %0 : vector<2xf32>
 }
 // CHECK-LABEL: @broadcast_vec1d_from_f32
 // CHECK-SAME:  %[[A:.*]]: f32)
-// CHECK:       %[[T0:.*]] = splat %[[A]] : vector<2xf32>
-// CHECK:       return %[[T0]] : vector<2xf32>
+// CHECK:       %[[T0:.*]] = llvm.insertelement %[[A]]
+// CHECK:       %[[T1:.*]] = llvm.shufflevector %[[T0]]
+// CHECK:       return %[[T1]] : vector<2xf32>
 
 // -----
 
 func @broadcast_vec1d_from_index(%arg0: index) -> vector<2xindex> {
   %0 = vector.broadcast %arg0 : index to vector<2xindex>
   return %0 : vector<2xindex>
 }
 // CHECK-LABEL: @broadcast_vec1d_from_index
 // CHECK-SAME:  %[[A:.*]]: index)
-// CHECK:       %[[T0:.*]] = splat %[[A]] : vector<2xindex>
-// CHECK:       return %[[T0]] : vector<2xindex>
+// CHECK:       %[[A1:.*]] = builtin.unrealized_conversion_cast %[[A]] : index to i64
+// CHECK:       %[[T0:.*]] = llvm.insertelement %[[A1]]
+// CHECK:       %[[T1:.*]] = llvm.shufflevector %[[T0]]
+// CHECK:       %[[T2:.*]] = builtin.unrealized_conversion_cast %[[T1]] : vector<2xi64> to vector<2xindex>
+// CHECK:       return %[[T2]] : vector<2xindex>
 
 // -----
 
 func @broadcast_vec2d_from_scalar(%arg0: f32) -> vector<2x3xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2x3xf32>
   return %0 : vector<2x3xf32>
 }
 // CHECK-LABEL: @broadcast_vec2d_from_scalar(
 // CHECK-SAME:  %[[A:.*]]: f32)
-// CHECK:       %[[T0:.*]] = splat %[[A]] : vector<2x3xf32>
-// CHECK:       return %[[T0]] : vector<2x3xf32>
+// CHECK:       %[[T0:.*]] = llvm.insertelement %[[A]]
+// CHECK:       %[[T1:.*]] = llvm.shufflevector %[[T0]]
+// CHECK:       %[[T2:.*]] = llvm.insertvalue %[[T1]], %{{.*}}[0] : !llvm.array<2 x vector<3xf32>> 
+// CHECK:       %[[T3:.*]] = llvm.insertvalue %[[T1]], %{{.*}}[1] : !llvm.array<2 x vector<3xf32>> 
+// CHECK:       %[[T4:.*]] = builtin.unrealized_conversion_cast %[[T3]] : !llvm.array<2 x vector<3xf32>> to vector<2x3xf32>
+// CHECK:       return %[[T4]] : vector<2x3xf32>
 
 // -----
 
 func @broadcast_vec3d_from_scalar(%arg0: f32) -> vector<2x3x4xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2x3x4xf32>
   return %0 : vector<2x3x4xf32>
 }
 // CHECK-LABEL: @broadcast_vec3d_from_scalar(
 // CHECK-SAME:  %[[A:.*]]: f32)
-// CHECK:       %[[T0:.*]] = splat %[[A]] : vector<2x3x4xf32>
-// CHECK:       return %[[T0]] : vector<2x3x4xf32>
+// CHECK:       %[[T0:.*]] = llvm.insertelement %[[A]]
+// CHECK:       %[[T1:.*]] = llvm.shufflevector %[[T0]]
+// CHECK:       %[[T2:.*]] = llvm.insertvalue %[[T1]], %{{.*}}[0, 0] : !llvm.array<2 x array<3 x vector<4xf32>>>
+// ...
+// CHECK:       %[[T3:.*]] = llvm.insertvalue %[[T1]], %{{.*}}[1, 2] : !llvm.array<2 x array<3 x vector<4xf32>>>
+// CHECK:       %[[T4:.*]] = builtin.unrealized_conversion_cast %[[T3]] : !llvm.array<2 x array<3 x vector<4xf32>>> to vector<2x3x4xf32>
+// CHECK:       return %[[T4]] : vector<2x3x4xf32>
 
 // -----
 
 func @broadcast_vec1d_from_vec1d(%arg0: vector<2xf32>) -> vector<2xf32> {
   %0 = vector.broadcast %arg0 : vector<2xf32> to vector<2xf32>
   return %0 : vector<2xf32>
 }
 // CHECK-LABEL: @broadcast_vec1d_from_vec1d(
 // CHECK-SAME:  %[[A:.*]]: vector<2xf32>)
 // CHECK:       return %[[A]] : vector<2xf32>
 
 // -----
 
 func @broadcast_vec2d_from_vec0d(%arg0: vector<f32>) -> vector<3x2xf32> {
   %0 = vector.broadcast %arg0 : vector<f32> to vector<3x2xf32>
   return %0 : vector<3x2xf32>
 }
 // CHECK-LABEL: @broadcast_vec2d_from_vec0d(
 // CHECK-SAME:  %[[A:.*]]: vector<f32>)
 //       CHECK: %[[T0:.*]] = builtin.unrealized_conversion_cast %[[A]] : vector<f32> to vector<1xf32>
 //       CHECK: %[[T1:.*]] = arith.constant dense<0.000000e+00> : vector<3x2xf32>
 //       CHECK: %[[T2:.*]] = builtin.unrealized_conversion_cast %[[T1]] : vector<3x2xf32> to !llvm.array<3 x vector<2xf32>>
 //       CHECK: %[[T4:.*]] = llvm.mlir.constant(0 : index) : i64
 //       CHECK: %[[T5:.*]] = llvm.extractelement %[[T0]][%[[T4]] : i64] : vector<1xf32>
-//       CHECK: %[[T6:.*]] = splat %[[T5]] : vector<2xf32>
+//       CHECK: %[[T6Insert:.*]] = llvm.insertelement %[[T5]]
+//       CHECK: %[[T6:.*]] = llvm.shufflevector %[[T6Insert]]
 //       CHECK: %[[T7:.*]] = llvm.insertvalue %[[T6]], %[[T2]][0] : !llvm.array<3 x vector<2xf32>>
 //       CHECK: %[[T8:.*]] = llvm.insertvalue %[[T6]], %[[T7]][1] : !llvm.array<3 x vector<2xf32>>
 //       CHECK: %[[T9:.*]] = llvm.insertvalue %[[T6]], %[[T8]][2] : !llvm.array<3 x vector<2xf32>>
 //       CHECK: %[[T10:.*]] = builtin.unrealized_conversion_cast %[[T9]] : !llvm.array<3 x vector<2xf32>> to vector<3x2xf32>
 //       CHECK: return %[[T10]] : vector<3x2xf32>
 
 // -----
 

 func @broadcast_stretch(%arg0: vector<1xf32>) -> vector<4xf32> {
   %0 = vector.broadcast %arg0 : vector<1xf32> to vector<4xf32>
   return %0 : vector<4xf32>
 }
 // CHECK-LABEL: @broadcast_stretch(
 // CHECK-SAME:  %[[A:.*]]: vector<1xf32>)
 // CHECK:       %[[T1:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T2:.*]] = llvm.extractelement %[[A]]{{\[}}%[[T1]] : i64] : vector<1xf32>
-// CHECK:       %[[T3:.*]] = splat %[[T2]] : vector<4xf32>
-// CHECK:       return %[[T3]] : vector<4xf32>
+// CHECK:       %[[T3:.*]] = llvm.insertelement %[[T2]]
+// CHECK:       %[[T4:.*]] = llvm.shufflevector %[[T3]]
+// CHECK:       return %[[T4]] : vector<4xf32>
 
 // -----
 
 func @broadcast_stretch_at_start(%arg0: vector<1x4xf32>) -> vector<3x4xf32> {
   %0 = vector.broadcast %arg0 : vector<1x4xf32> to vector<3x4xf32>
   return %0 : vector<3x4xf32>
 }
 // CHECK-LABEL: @broadcast_stretch_at_start(

 // CHECK-LABEL: @broadcast_stretch_at_end(
 // CHECK-SAME:  %[[A:.*]]: vector<4x1xf32>)
 // CHECK:       %[[T2:.*]] = builtin.unrealized_conversion_cast %[[A]] : vector<4x1xf32> to !llvm.array<4 x vector<1xf32>>
 // CHECK:       %[[T1:.*]] = arith.constant dense<0.000000e+00> : vector<4x3xf32>
 // CHECK:       %[[T7:.*]] = builtin.unrealized_conversion_cast %[[T1]] : vector<4x3xf32> to !llvm.array<4 x vector<3xf32>>
 // CHECK:       %[[T3:.*]] = llvm.extractvalue %[[T2]][0] : !llvm.array<4 x vector<1xf32>>
 // CHECK:       %[[T4:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T5:.*]] = llvm.extractelement %[[T3]]{{\[}}%[[T4]] : i64] : vector<1xf32>
-// CHECK:       %[[T6:.*]] = splat %[[T5]] : vector<3xf32>
+// CHECK:       %[[T6Insert:.*]] = llvm.insertelement %[[T5]]
+// CHECK:       %[[T6:.*]] = llvm.shufflevector %[[T6Insert]]
 // CHECK:       %[[T8:.*]] = llvm.insertvalue %[[T6]], %[[T7]][0] : !llvm.array<4 x vector<3xf32>>
 // CHECK:       %[[T10:.*]] = llvm.extractvalue %[[T2]][1] : !llvm.array<4 x vector<1xf32>>
 // CHECK:       %[[T11:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T12:.*]] = llvm.extractelement %[[T10]]{{\[}}%[[T11]] : i64] : vector<1xf32>
-// CHECK:       %[[T13:.*]] = splat %[[T12]] : vector<3xf32>
+// CHECK:       %[[T13Insert:.*]] = llvm.insertelement %[[T12]]
+// CHECK:       %[[T13:.*]] = llvm.shufflevector %[[T13Insert]]
 // CHECK:       %[[T14:.*]] = llvm.insertvalue %[[T13]], %[[T8]][1] : !llvm.array<4 x vector<3xf32>>
 // CHECK:       %[[T16:.*]] = llvm.extractvalue %[[T2]][2] : !llvm.array<4 x vector<1xf32>>
 // CHECK:       %[[T17:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T18:.*]] = llvm.extractelement %[[T16]]{{\[}}%[[T17]] : i64] : vector<1xf32>
-// CHECK:       %[[T19:.*]] = splat %[[T18]] : vector<3xf32>
+// CHECK:       %[[T19Insert:.*]] = llvm.insertelement %[[T18]]
+// CHECK:       %[[T19:.*]] = llvm.shufflevector %[[T19Insert]]
 // CHECK:       %[[T20:.*]] = llvm.insertvalue %[[T19]], %[[T14]][2] : !llvm.array<4 x vector<3xf32>>
 // CHECK:       %[[T22:.*]] = llvm.extractvalue %[[T2]][3] : !llvm.array<4 x vector<1xf32>>
 // CHECK:       %[[T23:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T24:.*]] = llvm.extractelement %[[T22]]{{\[}}%[[T23]] : i64] : vector<1xf32>
-// CHECK:       %[[T25:.*]] = splat %[[T24]] : vector<3xf32>
+// CHECK:       %[[T25Insert:.*]] = llvm.insertelement %[[T24]]
+// CHECK:       %[[T25:.*]] = llvm.shufflevector %[[T25Insert]]
 // CHECK:       %[[T26:.*]] = llvm.insertvalue %[[T25]], %[[T20]][3] : !llvm.array<4 x vector<3xf32>>
 // CHECK:       %[[T27:.*]] = builtin.unrealized_conversion_cast %[[T26]] : !llvm.array<4 x vector<3xf32>> to vector<4x3xf32>
 // CHECK:       return %[[T27]] : vector<4x3xf32>
 
 // -----
 
 func @broadcast_stretch_in_middle(%arg0: vector<4x1x2xf32>) -> vector<4x3x2xf32> {
   %0 = vector.broadcast %arg0 : vector<4x1x2xf32> to vector<4x3x2xf32>

 }
 // CHECK-LABEL: @outerproduct(
 // CHECK-SAME:  %[[A:.*]]: vector<2xf32>,
 // CHECK-SAME:  %[[B:.*]]: vector<3xf32>)
 // CHECK:       %[[T2:.*]] = arith.constant dense<0.000000e+00> : vector<2x3xf32>
 // CHECK:       %[[T7:.*]] = builtin.unrealized_conversion_cast %[[T2]] : vector<2x3xf32> to !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T3:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T4:.*]] = llvm.extractelement %[[A]]{{\[}}%[[T3]] : i64] : vector<2xf32>
-// CHECK:       %[[T5:.*]] = splat %[[T4]] : vector<3xf32>
+// CHECK:       %[[T5Insert:.*]] = llvm.insertelement %[[T4]]
+// CHECK:       %[[T5:.*]] = llvm.shufflevector %[[T5Insert]]
 // CHECK:       %[[T6:.*]] = arith.mulf %[[T5]], %[[B]] : vector<3xf32>
 // CHECK:       %[[T8:.*]] = llvm.insertvalue %[[T6]], %[[T7]][0] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T9:.*]] = llvm.mlir.constant(1 : i64) : i64
 // CHECK:       %[[T10:.*]] = llvm.extractelement %[[A]]{{\[}}%[[T9]] : i64] : vector<2xf32>
-// CHECK:       %[[T11:.*]] = splat %[[T10]] : vector<3xf32>
+// CHECK:       %[[T11Insert:.*]] = llvm.insertelement %[[T10]]
+// CHECK:       %[[T11:.*]] = llvm.shufflevector %[[T11Insert]]
 // CHECK:       %[[T12:.*]] = arith.mulf %[[T11]], %[[B]] : vector<3xf32>
 // CHECK:       %[[T13:.*]] = llvm.insertvalue %[[T12]], %[[T8]][1] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T14:.*]] = builtin.unrealized_conversion_cast %[[T13]] : !llvm.array<2 x vector<3xf32>> to vector<2x3xf32>
 // CHECK:       return %[[T14]] : vector<2x3xf32>
 
 // -----
 
 func @outerproduct_index(%arg0: vector<2xindex>, %arg1: vector<3xindex>) -> vector<2x3xindex> {
   %2 = vector.outerproduct %arg0, %arg1 : vector<2xindex>, vector<3xindex>
   return %2 : vector<2x3xindex>
 }
 // CHECK-LABEL: @outerproduct_index(
 // CHECK-SAME:  %[[A:.*]]: vector<2xindex>,
 // CHECK-SAME:  %[[B:.*]]: vector<3xindex>)
 // CHECK:       %[[T1:.*]] = builtin.unrealized_conversion_cast %[[A]] : vector<2xindex> to vector<2xi64>
 // CHECK:       %[[T0:.*]] = arith.constant dense<0> : vector<2x3xindex>
 // CHECK:       %[[T8:.*]] = builtin.unrealized_conversion_cast %[[T0]] : vector<2x3xindex> to !llvm.array<2 x vector<3xi64>>
 // CHECK:       %[[T2:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T3:.*]] = llvm.extractelement %[[T1]]{{\[}}%[[T2]] : i64] : vector<2xi64>
-// CHECK:       %[[T4:.*]] = builtin.unrealized_conversion_cast %[[T3]] : i64 to index
-// CHECK:       %[[T5:.*]] = splat %[[T4]] : vector<3xindex>
-// CHECK:       %[[T6:.*]] = arith.muli %[[T5]], %[[B]] : vector<3xindex>
+// CHECK:       %[[T4:.*]] = llvm.insertelement %[[T3]]
+// CHECK:       %[[T5:.*]] = llvm.shufflevector %[[T4]]
+// CHECK:       %[[T5Cast:.*]] = builtin.unrealized_conversion_cast %[[T5]] : vector<3xi64> to vector<3xindex>
+// CHECK:       %[[T6:.*]] = arith.muli %[[T5Cast]], %[[B]] : vector<3xindex>
 // CHECK:       %[[T7:.*]] = builtin.unrealized_conversion_cast %[[T6]] : vector<3xindex> to vector<3xi64>
 // CHECK:       %{{.*}} = llvm.insertvalue %[[T7]], %[[T8]][0] : !llvm.array<2 x vector<3xi64>>
 
 // -----
 
 func @outerproduct_add(%arg0: vector<2xf32>, %arg1: vector<3xf32>, %arg2: vector<2x3xf32>) -> vector<2x3xf32> {
   %2 = vector.outerproduct %arg0, %arg1, %arg2 : vector<2xf32>, vector<3xf32>
   return %2 : vector<2x3xf32>
 }
 // CHECK-LABEL: @outerproduct_add(
 // CHECK-SAME:  %[[A:.*]]: vector<2xf32>,
 // CHECK-SAME:  %[[B:.*]]: vector<3xf32>,
 // CHECK-SAME:  %[[C:.*]]: vector<2x3xf32>) -> vector<2x3xf32>
 // CHECK:       %[[T7:.*]] = builtin.unrealized_conversion_cast %[[C]] : vector<2x3xf32> to !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T3:.*]] = arith.constant dense<0.000000e+00> : vector<2x3xf32>
 // CHECK:       %[[T10:.*]] = builtin.unrealized_conversion_cast %[[T3]] : vector<2x3xf32> to !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T4:.*]] = llvm.mlir.constant(0 : i64) : i64
 // CHECK:       %[[T5:.*]] = llvm.extractelement %[[A]]{{\[}}%[[T4]] : i64] : vector<2xf32>
-// CHECK:       %[[T6:.*]] = splat %[[T5]] : vector<3xf32>
+// CHECK:       %[[T6Insert:.*]] = llvm.insertelement %[[T5]]
+// CHECK:       %[[T6:.*]] = llvm.shufflevector %[[T6Insert]]
 // CHECK:       %[[T8:.*]] = llvm.extractvalue %[[T7]][0] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T9:.*]] = "llvm.intr.fmuladd"(%[[T6]], %[[B]], %[[T8]]) : (vector<3xf32>, vector<3xf32>, vector<3xf32>) -> vector<3xf32>
 // CHECK:       %[[T11:.*]] = llvm.insertvalue %[[T9]], %[[T10]][0] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T12:.*]] = llvm.mlir.constant(1 : i64) : i64
 // CHECK:       %[[T13:.*]] = llvm.extractelement %[[A]]{{\[}}%[[T12]] : i64] : vector<2xf32>
-// CHECK:       %[[T14:.*]] = splat %[[T13]] : vector<3xf32>
+// CHECK:       %[[T14Insert:.*]] = llvm.insertelement %[[T13]]
+// CHECK:       %[[T14:.*]] = llvm.shufflevector %[[T14Insert]]
 // CHECK:       %[[T16:.*]] = llvm.extractvalue %[[T7]][1] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T17:.*]] = "llvm.intr.fmuladd"(%[[T14]], %[[B]], %[[T16]]) : (vector<3xf32>, vector<3xf32>, vector<3xf32>) -> vector<3xf32>
 // CHECK:       %[[T18:.*]] = llvm.insertvalue %[[T17]], %[[T11]][1] : !llvm.array<2 x vector<3xf32>>
 // CHECK:       %[[T19:.*]] = builtin.unrealized_conversion_cast %[[T18]] : !llvm.array<2 x vector<3xf32>> to vector<2x3xf32>
 // CHECK:       return %[[T19]] : vector<2x3xf32>
 
 // -----
 

 
 func @extract_strided_slice3(%arg0: vector<4x8xf32>) -> vector<2x2xf32> {
   %0 = vector.extract_strided_slice %arg0 {offsets = [2, 2], sizes = [2, 2], strides = [1, 1]} : vector<4x8xf32> to vector<2x2xf32>
   return %0 : vector<2x2xf32>
 }
 // CHECK-LABEL: @extract_strided_slice3(
 //  CHECK-SAME:    %[[ARG:.*]]: vector<4x8xf32>)
 //       CHECK:    %[[A:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : vector<4x8xf32> to !llvm.array<4 x vector<8xf32>>
-//       CHECK:    %[[VAL_1:.*]] = arith.constant 0.000000e+00 : f32
-//       CHECK:    %[[VAL_2:.*]] = splat %[[VAL_1]] : vector<2x2xf32>
+//       CHECK:    %[[VAL_2:.*]] = arith.constant dense<0.000000e+00> : vector<2x2xf32>
 //       CHECK:    %[[VAL_6:.*]] = builtin.unrealized_conversion_cast %[[VAL_2]] : vector<2x2xf32> to !llvm.array<2 x vector<2xf32>>
 //       CHECK:    %[[T2:.*]] = llvm.extractvalue %[[A]][2] : !llvm.array<4 x vector<8xf32>>
 //       CHECK:    %[[T3:.*]] = llvm.shufflevector %[[T2]], %[[T2]] [2, 3] : vector<8xf32>, vector<8xf32>
 //       CHECK:    %[[T4:.*]] = llvm.insertvalue %[[T3]], %[[VAL_6]][0] : !llvm.array<2 x vector<2xf32>>
 //       CHECK:    %[[T5:.*]] = llvm.extractvalue %[[A]][3] : !llvm.array<4 x vector<8xf32>>
 //       CHECK:    %[[T6:.*]] = llvm.shufflevector %[[T5]], %[[T5]] [2, 3] : vector<8xf32>, vector<8xf32>
 //       CHECK:    %[[T7:.*]] = llvm.insertvalue %[[T6]], %[[T4]][1] : !llvm.array<2 x vector<2xf32>>
 //       CHECK:    %[[VAL_12:.*]] = builtin.unrealized_conversion_cast %[[T7]] : !llvm.array<2 x vector<2xf32>> to vector<2x2xf32>

 //
 // 1. Create a vector with linear indices [ 0 .. vector_length - 1 ].
 //       CHECK: %[[linearIndex:.*]] = arith.constant dense
 //  CHECK-SAME: <[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]> :
 //  CHECK-SAME: vector<17xi32>
 //
 // 2. Create offsetVector = [ offset + 0 .. offset + vector_length - 1 ].
 //       CHECK: %[[otrunc:.*]] = arith.index_cast %[[BASE]] : index to i32
-//       CHECK: %[[offsetVec:.*]] = splat %[[otrunc]] : vector<17xi32>
+//       CHECK: %[[offsetVecInsert:.*]] = llvm.insertelement %[[otrunc]]
+//       CHECK: %[[offsetVec:.*]] = llvm.shufflevector %[[offsetVecInsert]]
 //       CHECK: %[[offsetVec2:.*]] = arith.addi %[[offsetVec]], %[[linearIndex]] : vector<17xi32>
 //
 // 3. Let dim the memref dimension, compute the vector comparison mask:
 //    [ offset + 0 .. offset + vector_length - 1 ] < [ dim .. dim ]
 //       CHECK: %[[dtrunc:.*]] = arith.index_cast %[[DIM]] : index to i32
-//       CHECK: %[[dimVec:.*]] = splat %[[dtrunc]] : vector<17xi32>
+//       CHECK: %[[dimVecInsert:.*]] = llvm.insertelement %[[dtrunc]]
+//       CHECK: %[[dimVec:.*]] = llvm.shufflevector %[[dimVecInsert]]
 //       CHECK: %[[mask:.*]] = arith.cmpi slt, %[[offsetVec2]], %[[dimVec]] : vector<17xi32>
 //
 // 4. Create pass-through vector.
-//       CHECK: %[[PASS_THROUGH:.*]] = splat %[[c7]] : vector<17xf32>
+//       CHECK: %[[PASS_THROUGH:.*]] = arith.constant dense<7.{{.*}}> : vector<17xf32>
 //
 // 5. Bitcast to vector form.
 //       CHECK: %[[gep:.*]] = llvm.getelementptr {{.*}} :
 //  CHECK-SAME: (!llvm.ptr<f32>, i64) -> !llvm.ptr<f32>
 //       CHECK: %[[vecPtr:.*]] = llvm.bitcast %[[gep]] :
 //  CHECK-SAME: !llvm.ptr<f32> to !llvm.ptr<vector<17xf32>>
 //
 // 6. Rewrite as a masked read.
 //       CHECK: %[[loaded:.*]] = llvm.intr.masked.load %[[vecPtr]], %[[mask]],
 //  CHECK-SAME: %[[PASS_THROUGH]] {alignment = 4 : i32} :
 //  CHECK-SAME: (!llvm.ptr<vector<17xf32>>, vector<17xi1>, vector<17xf32>) -> vector<17xf32>
 //
 // 1. Create a vector with linear indices [ 0 .. vector_length - 1 ].
 //       CHECK: %[[linearIndex_b:.*]] = arith.constant dense
 //  CHECK-SAME: <[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]> :
 //  CHECK-SAME: vector<17xi32>
 //
 // 2. Create offsetVector = [ offset + 0 .. offset + vector_length - 1 ].
-//       CHECK: splat %{{.*}} : vector<17xi32>
+//       CHECK: llvm.shufflevector %{{.*}} : vector<17xi32>
 //       CHECK: arith.addi
 //
 // 3. Let dim the memref dimension, compute the vector comparison mask:
 //    [ offset + 0 .. offset + vector_length - 1 ] < [ dim .. dim ]
-//       CHECK: splat %{{.*}} : vector<17xi32>
+//       CHECK: llvm.shufflevector %{{.*}} : vector<17xi32>
 //       CHECK: %[[mask_b:.*]] = arith.cmpi slt, {{.*}} : vector<17xi32>
 //
 // 4. Bitcast to vector form.
 //       CHECK: %[[gep_b:.*]] = llvm.getelementptr {{.*}} :
 //  CHECK-SAME: (!llvm.ptr<f32>, i64) -> !llvm.ptr<f32>
 //       CHECK: %[[vecPtr_b:.*]] = llvm.bitcast %[[gep_b]] :
 //  CHECK-SAME: !llvm.ptr<f32> to !llvm.ptr<vector<17xf32>>
 //
%f = vector.transfer_read %A[%base], %f7
     memref<?xindex>, vector<17xindex>
   vector.transfer_write %f, %A[%base]
       {permutation_map = affine_map<(d0) -> (d0)>} :
     vector<17xindex>, memref<?xindex>
   return %f: vector<17xindex>
 }
 // CHECK-LABEL: func @transfer_read_index_1d
 //  CHECK-SAME: %[[BASE:[a-zA-Z0-9]*]]: index) -> vector<17xindex>
-//       CHECK: %[[C7:.*]] = arith.constant 7 : index
-//       CHECK: %[[SPLAT:.*]] = splat %[[C7]] : vector<17xindex>
+//       CHECK: %[[SPLAT:.*]] = arith.constant dense<7> : vector<17xindex>
 //       CHECK: %{{.*}} = builtin.unrealized_conversion_cast %[[SPLAT]] : vector<17xindex> to vector<17xi64>
 
 //       CHECK: %[[loaded:.*]] = llvm.intr.masked.load %{{.*}}, %{{.*}}, %{{.*}} {alignment = 8 : i32} :
 //  CHECK-SAME: (!llvm.ptr<vector<17xi64>>, vector<17xi1>, vector<17xi64>) -> vector<17xi64>
 
 //       CHECK: llvm.intr.masked.store %[[loaded]], %{{.*}}, %{{.*}} {alignment = 8 : i32} :
 //  CHECK-SAME: vector<17xi64>, vector<17xi1> into !llvm.ptr<vector<17xi64>>
 
 // -----
 
 func @transfer_read_2d_to_1d(%A : memref<?x?xf32>, %base0: index, %base1: index) -> vector<17xf32> {
   %f7 = arith.constant 7.0: f32
   %f = vector.transfer_read %A[%base0, %base1], %f7
       {permutation_map = affine_map<(d0, d1) -> (d1)>} :
     memref<?x?xf32>, vector<17xf32>
   return %f: vector<17xf32>
 }
 // CHECK-LABEL: func @transfer_read_2d_to_1d
 //  CHECK-SAME: %[[BASE_0:[a-zA-Z0-9]*]]: index, %[[BASE_1:[a-zA-Z0-9]*]]: index) -> vector<17xf32>
 //       CHECK: %[[c1:.*]] = arith.constant 1 : index
 //       CHECK: %[[DIM:.*]] = memref.dim %{{.*}}, %[[c1]] : memref<?x?xf32>
 //
 // Create offsetVector = [ offset + 0 .. offset + vector_length - 1 ].
 //       CHECK: %[[trunc:.*]] = arith.index_cast %[[BASE_1]] : index to i32
-//       CHECK: %[[offsetVec:.*]] = splat %[[trunc]] : vector<17xi32>
+//       CHECK: %[[offsetVecInsert:.*]] = llvm.insertelement %[[trunc]]
+//       CHECK: %[[offsetVec:.*]] = llvm.shufflevector %[[offsetVecInsert]]
 //
 // Let dim the memref dimension, compute the vector comparison mask:
 //    [ offset + 0 .. offset + vector_length - 1 ] < [ dim .. dim ]
 //       CHECK: %[[dimtrunc:.*]] = arith.index_cast %[[DIM]] : index to i32
-//       CHECK: splat %[[dimtrunc]] : vector<17xi32>
+//       CHECK: %[[dimtruncInsert:.*]] = llvm.insertelement %[[dimtrunc]]
+//       CHECK: llvm.shufflevector %[[dimtruncInsert]]
 
 // -----
 
 func @transfer_read_1d_non_zero_addrspace(%A : memref<?xf32, 3>, %base: index) -> vector<17xf32> {
   %f7 = arith.constant 7.0: f32
   %f = vector.transfer_read %A[%base], %f7
       {permutation_map = affine_map<(d0) -> (d0)>} :
     memref<?xf32, 3>, vector<17xf32>
func @create_mask_0d(%a : index) -> vector<i1> {
   %v = vector.create_mask %a : vector<i1>
   return %v: vector<i1>
 }
 
 // CHECK-LABEL: func @create_mask_0d
 // CHECK-SAME: %[[arg:.*]]: index
 // CHECK:  %[[indices:.*]] = arith.constant dense<0> : vector<i32>
 // CHECK:  %[[arg_i32:.*]] = arith.index_cast %[[arg]] : index to i32
-// CHECK:  %[[bounds:.*]] = splat %[[arg_i32]] : vector<i32>
-// CHECK:  %[[result:.*]] = arith.cmpi slt, %[[indices]], %[[bounds]] : vector<i32>
+// CHECK:  %[[bounds:.*]] = llvm.insertelement %[[arg_i32]]
+// CHECK:  %[[boundsCast:.*]] = builtin.unrealized_conversion_cast %[[bounds]] : vector<1xi32> to vector<i32>
+// CHECK:  %[[result:.*]] = arith.cmpi slt, %[[indices]], %[[boundsCast]] : vector<i32>
 // CHECK:  return %[[result]] : vector<i1>
+
 // -----
 
 func @create_mask_1d(%a : index) -> vector<4xi1> {
   %v = vector.create_mask %a : vector<4xi1>
   return %v: vector<4xi1>
 }
 
 // CHECK-LABEL: func @create_mask_1d
 // CHECK-SAME: %[[arg:.*]]: index
 // CHECK:  %[[indices:.*]] = arith.constant dense<[0, 1, 2, 3]> : vector<4xi32>
 // CHECK:  %[[arg_i32:.*]] = arith.index_cast %[[arg]] : index to i32
-// CHECK:  %[[bounds:.*]] = splat %[[arg_i32]] : vector<4xi32>
+// CHECK:  %[[boundsInsert:.*]] = llvm.insertelement %[[arg_i32]]
+// CHECK:  %[[bounds:.*]] = llvm.shufflevector %[[boundsInsert]]
 // CHECK:  %[[result:.*]] = arith.cmpi slt, %[[indices]], %[[bounds]] : vector<4xi32>
 // CHECK:  return %[[result]] : vector<4xi1>
 
 // -----
 
 func @flat_transpose(%arg0: vector<16xf32>) -> vector<16xf32> {
   %0 = vector.flat_transpose %arg0 { rows = 4: i32, columns = 4: i32 }
      : vector<16xf32> -> vector<16xf32>

 
 func @compress_store_op_index(%arg0: memref<?xindex>, %arg1: vector<11xi1>, %arg2: vector<11xindex>) {
   %c0 = arith.constant 0: index
   vector.compressstore %arg0[%c0], %arg1, %arg2 : memref<?xindex>, vector<11xi1>, vector<11xindex>
   return
 }
 // CHECK-LABEL: func @compress_store_op_index
 // CHECK: "llvm.intr.masked.compressstore"(%{{.*}}, %{{.*}}, %{{.*}}) : (vector<11xi64>, !llvm.ptr<i64>, vector<11xi1>) -> ()
+
+// -----
+
+// CHECK-LABEL: @splat_0d
+// CHECK-SAME: %[[ARG:.*]]: f32
+func @splat_0d(%a: f32) -> vector<f32> {
+  %v = vector.splat %a : vector<f32>
+  return %v : vector<f32>
+}
+// CHECK-NEXT: %[[UNDEF:[0-9]+]] = llvm.mlir.undef : vector<1xf32>
+// CHECK-NEXT: %[[ZERO:[0-9]+]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK-NEXT: %[[V:[0-9]+]] = llvm.insertelement %[[ARG]], %[[UNDEF]][%[[ZERO]] : i32] : vector<1xf32>
+// CHECK-NEXT: %[[VCAST:[0-9]+]] = builtin.unrealized_conversion_cast %[[V]] : vector<1xf32> to vector<f32>
+// CHECK-NEXT: return %[[VCAST]] : vector<f32>
+
+// -----
+
+// CHECK-LABEL: @splat
+// CHECK-SAME: %[[A:arg[0-9]+]]: vector<4xf32>
+// CHECK-SAME: %[[ELT:arg[0-9]+]]: f32
+func @splat(%a: vector<4xf32>, %b: f32) -> vector<4xf32> {
+  %vb = vector.splat %b : vector<4xf32>
+  %r = arith.mulf %a, %vb : vector<4xf32>
+  return %r : vector<4xf32>
+}
+// CHECK-NEXT: %[[UNDEF:[0-9]+]] = llvm.mlir.undef : vector<4xf32>
+// CHECK-NEXT: %[[ZERO:[0-9]+]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK-NEXT: %[[V:[0-9]+]] = llvm.insertelement %[[ELT]], %[[UNDEF]][%[[ZERO]] : i32] : vector<4xf32>
+// CHECK-NEXT: %[[SPLAT:[0-9]+]] = llvm.shufflevector %[[V]], %[[UNDEF]] [0 : i32, 0 : i32, 0 : i32, 0 : i32]
+// CHECK-NEXT: %[[SCALE:[0-9]+]] = arith.mulf %[[A]], %[[SPLAT]] : vector<4xf32>
+// CHECK-NEXT: return %[[SCALE]] : vector<4xf32>

mlir/test/Conversion/VectorToSPIRV/simple.mlir

 
 // CHECK-LABEL: @fma
 //  CHECK-SAME: %[[A:.*]]: vector<4xf32>, %[[B:.*]]: vector<4xf32>, %[[C:.*]]: vector<4xf32>
 //       CHECK:   spv.GLSL.Fma %[[A]], %[[B]], %[[C]] : vector<4xf32>
 func @fma(%a: vector<4xf32>, %b: vector<4xf32>, %c: vector<4xf32>) -> vector<4xf32> {
   %0 = vector.fma %a, %b, %c: vector<4xf32>
   return %0 : vector<4xf32>
 }
+
+// -----
+
+// CHECK-LABEL: func @splat
+//  CHECK-SAME: (%[[A:.+]]: f32)
+//       CHECK:   %[[VAL:.+]] = spv.CompositeConstruct %[[A]], %[[A]], %[[A]], %[[A]] : vector<4xf32>
+//       CHECK:   return %[[VAL]]
+func @splat(%f : f32) -> vector<4xf32> {
+  %splat = vector.splat %f : vector<4xf32>
+  return %splat : vector<4xf32>
+}

mlir/test/Dialect/Standard/ops.mlir

func @switch_i64(%flag : i64, %caseOperand : i32) {
 
   ^bb1(%bb1arg : i32):
     return
   ^bb2(%bb2arg : i32):
     return
   ^bb3(%bb3arg : i32):
     return
 }
-
-// CHECK-LABEL: func @vector_splat_0d(
-func @vector_splat_0d(%a: f32) -> vector<f32> {
-  // CHECK: splat %{{.*}} : vector<f32>
-  %0 = splat %a : vector<f32>
-  return %0 : vector<f32>
-}

mlir/test/Dialect/Tensor/canonicalize.mlir

func @propogate_index_cast(%arg0: tensor<1xi32>) -> index {
   // CHECK: %[[EXT:.+]] = tensor.extract %arg0[%[[IDX]]] : tensor<1xi32>
   // CHECK: %[[CAST:.+]] = arith.index_cast %[[EXT]]
   // CHECK: return %[[CAST]] : index
   %c0 = arith.constant 0 : index
   %0 = arith.index_cast %arg0 : tensor<1xi32> to tensor<1xindex>
   %1 = tensor.extract %0[%c0] : tensor<1xindex>
   return %1 : index
 }
+
+// -----
+
+// CHECK-LABEL: func @splat_fold
+func @splat_fold() -> tensor<4xf32> {
+  %c = arith.constant 1.0 : f32
+  %t = tensor.splat %c : tensor<4xf32>
+  return %t : tensor<4xf32>
+
+  // CHECK-NEXT: [[T:%.*]] = arith.constant dense<1.000000e+00> : tensor<4xf32>
+  // CHECK-NEXT: return [[T]] : tensor<4xf32>
+}

mlir/test/Dialect/Tensor/invalid.mlir

func @pad_yield_type(%arg0: tensor<?x4xi32>, %arg1: i8) -> tensor<?x9xi32> {
   // expected-error @+1 {{op expected yield type to match shape element type}}
   %0 = tensor.pad %arg0 low[1, 2] high[2, 3] {
   ^bb0(%arg2: index, %arg3: index):
     tensor.yield %arg1 : i8
   } : tensor<?x4xi32> to tensor<?x9xi32>
   return %0 : tensor<?x9xi32>
 }
 
+// -----
+
+func @invalid_splat(%v : f32) {
+  // expected-error@+1 {{invalid kind of type specified}}
+  tensor.splat %v : memref<8xf32>
+  return
+}
+
+// -----
+
+func @invalid_splat(%v : vector<8xf32>) {
+  // expected-error@+1 {{must be integer/index/float type}}
+  %w = tensor.splat %v : tensor<8xvector<8xf32>>
+  return
+}
 No newline at end of file

mlir/test/Dialect/Tensor/ops.mlir

 //  CHECK-SAME: %[[UB1:[a-zA-Z0-9_]*]]
 //       CHECK:   tensor.pad %[[ARG0]]
 //  CHECK-SAME:     low[0, 0]
 //  CHECK-SAME:     high[%[[UB0]], %[[UB1]]]
 //       CHECK:    : tensor<?x?xf32> to tensor<2x3xf32>
 
 // -----
 
+// CHECK-LABEL: func @test_splat_op
+// CHECK-SAME: [[S:%arg[0-9]+]]: f32
+func @test_splat_op(%s : f32) {
+  // CHECK: tensor.splat [[S]] : tensor<8xf32>
+  %v = tensor.splat %s : tensor<8xf32>
+  
+  // CHECK: tensor.splat [[S]] : tensor<4xf32>
+  %u = "tensor.splat"(%s) : (f32) -> tensor<4xf32>
+  return
+}

mlir/test/Dialect/Vector/canonicalize.mlir

 }
 
 // -----
 
 // CHECK-LABEL: fold_extract_splat
 //  CHECK-SAME:   %[[A:.*]]: f32
 //       CHECK:   return %[[A]] : f32
 func @fold_extract_splat(%a : f32) -> f32 {
-  %b = splat %a : vector<1x2x4xf32>
+  %b = vector.splat %a : vector<1x2x4xf32>
   %r = vector.extract %b[0, 1, 2] : vector<1x2x4xf32>
   return %r : f32
 }
 
 // -----
 
 // CHECK-LABEL: fold_extract_broadcast_vector
 //  CHECK-SAME:   %[[A:.*]]: vector<4xf32>
func @insert_strided_slice_full_range(%source: vector<16x16xf16>, %dest: vector<16x16xf16>) -> vector<16x16xf16> {
   %0 = vector.insert_strided_slice %source, %dest {offsets = [0, 0], strides = [1, 1]} : vector<16x16xf16> into vector<16x16xf16>
   // CHECK: return %[[SOURCE]]
   return %0: vector<16x16xf16>
 }
 
 // -----
 
 // CHECK-LABEL: extract_strided_splat
-//       CHECK:   %[[B:.*]] = splat %{{.*}} : vector<2x4xf16>
+//       CHECK:   %[[B:.*]] = vector.splat %{{.*}} : vector<2x4xf16>
 //  CHECK-NEXT:   return %[[B]] : vector<2x4xf16>
 func @extract_strided_splat(%arg0: f16) -> vector<2x4xf16> {
- %0 = splat %arg0 : vector<16x4xf16>
+ %0 = vector.splat %arg0 : vector<16x4xf16>
  %1 = vector.extract_strided_slice %0
   {offsets = [1, 0], sizes = [2, 4], strides = [1, 1]} :
   vector<16x4xf16> to vector<2x4xf16>
   return %1 : vector<2x4xf16>
 }
 
 // -----
 

 //       CHECK: %[[V:.*]] = vector.extract %[[A]][1] : vector<2x4xf32>
 //       CHECK: return %[[V]] : vector<4xf32>
 func @extract_extract_strided2(%A: vector<2x4xf32>)
   -> (vector<4xf32>) {
  %0 = vector.extract_strided_slice %A {offsets = [1, 0], sizes = [1, 4], strides = [1, 1]} : vector<2x4xf32> to vector<1x4xf32>
  %1 = vector.extract %0[0] : vector<1x4xf32>
  return %1 : vector<4xf32>
 }
+
+// -----
+
+// CHECK-LABEL: func @splat_fold
+func @splat_fold() -> vector<4xf32> {
+  %c = arith.constant 1.0 : f32
+  %v = vector.splat %c : vector<4xf32>
+  return %v : vector<4xf32>
+
+  // CHECK-NEXT: [[V:%.*]] = arith.constant dense<1.000000e+00> : vector<4xf32>
+  // CHECK-NEXT: return [[V]] : vector<4xf32>
+}

mlir/test/Dialect/Vector/invalid.mlir

func @test_vector.transfer_read(%arg0: memref<?x?xf32>) {
   %0 = vector.transfer_read %arg0[%c3, %c3], %cst { permutation_map = affine_map<()->(0)> } : memref<?x?xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: vector<4x3xf32>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4x3xf32>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
   // expected-error@+1 {{ requires memref or ranked tensor type}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 : vector<4x3xf32>, vector<1x1x2x3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<4x3xf32>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4x3xf32>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
   // expected-error@+1 {{ requires vector type}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 : memref<4x3xf32>, f32
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xf32>) {
   %c3 = arith.constant 3 : index

 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?x?xf32>) {
   %c1 = arith.constant 1 : i1
   %c3 = arith.constant 3 : index
   %cst = arith.constant 3.0 : f32
   // expected-note@+1 {{prior use here}}
-  %mask = splat %c1 : vector<3x8x7xi1>
+  %mask = vector.splat %c1 : vector<3x8x7xi1>
   // expected-error@+1 {{expects different type than prior uses: 'vector<3x7xi1>' vs 'vector<3x8x7xi1>'}}
   %0 = vector.transfer_read %arg0[%c3, %c3, %c3], %cst, %mask {permutation_map = affine_map<(d0, d1, d2)->(d0, 0, d2)>} : memref<?x?x?xf32>, vector<3x8x7xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xvector<4x3xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4x3xf32>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
   // expected-error@+1 {{requires source vector element and vector result ranks to match}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 {permutation_map = affine_map<(d0, d1)->(d0, d1)>} : memref<?x?xvector<4x3xf32>>, vector<3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xvector<6xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<6xf32>
+  %vf0 = vector.splat %f0 : vector<6xf32>
   // expected-error@+1 {{requires the bitwidth of the minor 1-D vector to be an integral multiple of the bitwidth of the minor 1-D vector of the source}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 : memref<?x?xvector<6xf32>>, vector<3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xvector<2x3xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<2x3xf32>
+  %vf0 = vector.splat %f0 : vector<2x3xf32>
   // expected-error@+1 {{ expects the optional in_bounds attr of same rank as permutation_map results: affine_map<(d0, d1) -> (d0, d1)>}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 {in_bounds = [true], permutation_map = affine_map<(d0, d1)->(d0, d1)>} : memref<?x?xvector<2x3xf32>>, vector<1x1x2x3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xvector<2x3xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<2x3xf32>
+  %vf0 = vector.splat %f0 : vector<2x3xf32>
   // expected-error@+1 {{requires broadcast dimensions to be in-bounds}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0 {in_bounds = [false, true], permutation_map = affine_map<(d0, d1)->(0, d1)>} : memref<?x?xvector<2x3xf32>>, vector<1x1x2x3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_read(%arg0: memref<?x?xvector<2x3xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<2x3xf32>
-  %mask = splat %c1 : vector<2x3xi1>
+  %vf0 = vector.splat %f0 : vector<2x3xf32>
+  %mask = vector.splat %c1 : vector<2x3xi1>
   // expected-error@+1 {{does not support masks with vector element type}}
   %0 = vector.transfer_read %arg0[%c3, %c3], %vf0, %mask {permutation_map = affine_map<(d0, d1)->(d0, d1)>} : memref<?x?xvector<2x3xf32>>, vector<1x1x2x3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_write(%arg0: memref<?x?xf32>) {
   %c3 = arith.constant 3 : index
   %cst = arith.constant 3.0 : f32
   // expected-error@+1 {{requires two types}}
   vector.transfer_write %arg0, %arg0[%c3, %c3] : memref<?x?xf32>
 }
 
 // -----
 
 func @test_vector.transfer_write(%arg0: memref<vector<4x3xf32>>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4x3xf32>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
   // expected-error@+1 {{ requires vector type}}
   vector.transfer_write %arg0, %arg0[%c3, %c3] : memref<vector<4x3xf32>>, vector<4x3xf32>
 }
 
 // -----
 
 func @test_vector.transfer_write(%arg0: vector<4x3xf32>) {
   %c3 = arith.constant 3 : index
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4x3xf32>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
   // expected-error@+1 {{ requires memref or ranked tensor type}}
   vector.transfer_write %arg0, %arg0[%c3, %c3] : vector<4x3xf32>, f32
 }
 
 // -----
 
 func @test_vector.transfer_write(%arg0: memref<?x?xf32>) {
   %c3 = arith.constant 3 : index

 // -----
 
 func @scan_incompatible_shapes(%arg0: vector<2x3xi32>, %arg1: vector<5xi32>) -> vector<2x3xi32> {
   // expected-error@+1 {{incompatible input/initial value shapes}}
   %0:2 = vector.scan <add>, %arg0, %arg1 {inclusive = true, reduction_dim = 0} :
     vector<2x3xi32>, vector<5xi32>
   return %0#0 : vector<2x3xi32>
 }
+
+// -----
+
+func @invalid_splat(%v : f32) {
+  // expected-error@+1 {{invalid kind of type specified}}
+  vector.splat %v : memref<8xf32>
+  return
+}

mlir/test/Dialect/Vector/ops.mlir

func @vector_transfer_ops(%arg0: memref<?x?xf32>,
   // CHECK: %[[C3:.*]] = arith.constant 3 : index
   %c3 = arith.constant 3 : index
   %cst = arith.constant 3.0 : f32
   %f0 = arith.constant 0.0 : f32
   %c0 = arith.constant 0 : i32
   %i0 = arith.constant 0 : index
   %i1 = arith.constant 1 : i1
 
-  %vf0 = splat %f0 : vector<4x3xf32>
-  %v0 = splat %c0 : vector<4x3xi32>
-  %vi0 = splat %i0 : vector<4x3xindex>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
+  %v0 = vector.splat %c0 : vector<4x3xi32>
+  %vi0 = vector.splat %i0 : vector<4x3xindex>
   %m = arith.constant dense<[0, 0, 1, 0, 1]> : vector<5xi1>
-  %m2 = splat %i1 : vector<5x4xi1>
+  %m2 = vector.splat %i1 : vector<5x4xi1>
   //
   // CHECK: vector.transfer_read
   %0 = vector.transfer_read %arg0[%c3, %c3], %f0 {permutation_map = affine_map<(d0, d1)->(d0)>} : memref<?x?xf32>, vector<128xf32>
   // CHECK: vector.transfer_read
   %1 = vector.transfer_read %arg0[%c3, %c3], %f0 {permutation_map = affine_map<(d0, d1)->(d1, d0)>} : memref<?x?xf32>, vector<3x7xf32>
   // CHECK: vector.transfer_read
   %2 = vector.transfer_read %arg0[%c3, %c3], %cst {permutation_map = affine_map<(d0, d1)->(d0)>} : memref<?x?xf32>,  vector<128xf32>
   // CHECK: vector.transfer_read
func @vector_transfer_ops_tensor(%arg0: tensor<?x?xf32>,
    tensor<?x?xvector<4x3xindex>>){
   // CHECK: %[[C3:.*]] = arith.constant 3 : index
   %c3 = arith.constant 3 : index
   %cst = arith.constant 3.0 : f32
   %f0 = arith.constant 0.0 : f32
   %c0 = arith.constant 0 : i32
   %i0 = arith.constant 0 : index
 
-  %vf0 = splat %f0 : vector<4x3xf32>
-  %v0 = splat %c0 : vector<4x3xi32>
-  %vi0 = splat %i0 : vector<4x3xindex>
+  %vf0 = vector.splat %f0 : vector<4x3xf32>
+  %v0 = vector.splat %c0 : vector<4x3xi32>
+  %vi0 = vector.splat %i0 : vector<4x3xindex>
 
   //
   // CHECK: vector.transfer_read
   %0 = vector.transfer_read %arg0[%c3, %c3], %f0 {permutation_map = affine_map<(d0, d1)->(d0)>} : tensor<?x?xf32>, vector<128xf32>
   // CHECK: vector.transfer_read
   %1 = vector.transfer_read %arg0[%c3, %c3], %f0 {permutation_map = affine_map<(d0, d1)->(d1, d0)>} : tensor<?x?xf32>, vector<3x7xf32>
   // CHECK: vector.transfer_read
   %2 = vector.transfer_read %arg0[%c3, %c3], %cst {permutation_map = affine_map<(d0, d1)->(d0)>} : tensor<?x?xf32>,  vector<128xf32>

 
 // CHECK-LABEL: @vector_scan
 func @vector_scan(%0: vector<4x8x16x32xf32>) -> vector<4x8x16x32xf32> {
   %1 = arith.constant dense<0.0> : vector<4x16x32xf32>
   %2:2 = vector.scan <add>, %0, %1 {reduction_dim = 1 : i64, inclusive = true} :
     vector<4x8x16x32xf32>, vector<4x16x32xf32>
   return %2#0 : vector<4x8x16x32xf32>
 }
+
+// CHECK-LABEL: func @test_splat_op
+// CHECK-SAME: [[S:%arg[0-9]+]]: f32
+func @test_splat_op(%s : f32) {
+  // CHECK: vector.splat [[S]] : vector<8xf32>
+  %v = vector.splat %s : vector<8xf32>
+  
+  // CHECK: vector.splat [[S]] : vector<4xf32>
+  %u = "vector.splat"(%s) : (f32) -> vector<4xf32>
+  return
+}
+
+// CHECK-LABEL: func @vector_splat_0d(
+func @vector_splat_0d(%a: f32) -> vector<f32> {
+  // CHECK: vector.splat %{{.*}} : vector<f32>
+  %0 = vector.splat %a : vector<f32>
+  return %0 : vector<f32>
+}

mlir/test/Dialect/Vector/vector-contract-transforms.mlir

func @full_contract2(%arg0: vector<2x3xf32>,
   return %0 : f32
 }
 
 // CHECK-LABEL: func @outerproduct_noacc
 // CHECK-SAME: %[[A:.*0]]: vector<2xf32>,
 // CHECK-SAME: %[[B:.*1]]: vector<3xf32>
 // CHECK:      %[[C0:.*]] = arith.constant dense<0.000000e+00> : vector<2x3xf32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : vector<2xf32>
-// CHECK:      %[[T1:.*]] = splat %[[T0]] : vector<3xf32>
+// CHECK:      %[[T1:.*]] = vector.splat %[[T0]] : vector<3xf32>
 // CHECK:      %[[T2:.*]] = arith.mulf %[[T1]], %[[B]] : vector<3xf32>
 // CHECK:      %[[T3:.*]] = vector.insert %[[T2]], %[[C0]] [0] : vector<3xf32> into vector<2x3xf32>
 // CHECK:      %[[T4:.*]] = vector.extract %[[A]][1] : vector<2xf32>
-// CHECK:      %[[T5:.*]] = splat %[[T4]] : vector<3xf32>
+// CHECK:      %[[T5:.*]] = vector.splat %[[T4]] : vector<3xf32>
 // CHECK:      %[[T6:.*]] = arith.mulf %[[T5]], %[[B]] : vector<3xf32>
 // CHECK:      %[[T7:.*]] = vector.insert %[[T6]], %[[T3]] [1] : vector<3xf32> into vector<2x3xf32>
 // CHECK:      return %[[T7]] : vector<2x3xf32>
 
 func @outerproduct_noacc(%arg0: vector<2xf32>,
                          %arg1: vector<3xf32>) -> vector<2x3xf32> {
   %0 = vector.outerproduct %arg0, %arg1 : vector<2xf32>, vector<3xf32>
   return %0: vector<2x3xf32>
 }
 
 // CHECK-LABEL: func @outerproduct_acc
 // CHECK-SAME: %[[A:.*0]]: vector<2xf32>,
 // CHECK-SAME: %[[B:.*1]]: vector<3xf32>,
 // CHECK-SAME: %[[C:.*2]]: vector<2x3xf32>
 // CHECK:      %[[C0:.*]] = arith.constant dense<0.000000e+00> : vector<2x3xf32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : vector<2xf32>
-// CHECK:      %[[T1:.*]] = splat %[[T0]] : vector<3xf32>
+// CHECK:      %[[T1:.*]] = vector.splat %[[T0]] : vector<3xf32>
 // CHECK:      %[[T2:.*]] = vector.extract %[[C]][0] : vector<2x3xf32>
 // CHECK:      %[[T3:.*]] = vector.fma %[[T1]], %[[B]], %[[T2]] : vector<3xf32>
 // CHECK:      %[[T4:.*]] = vector.insert %[[T3]], %[[C0]] [0] : vector<3xf32> into vector<2x3xf32>
 // CHECK:      %[[T5:.*]] = vector.extract %[[A]][1] : vector<2xf32>
-// CHECK:      %[[T6:.*]] = splat %[[T5]] : vector<3xf32>
+// CHECK:      %[[T6:.*]] = vector.splat %[[T5]] : vector<3xf32>
 // CHECK:      %[[T7:.*]] = vector.extract %[[C]][1] : vector<2x3xf32>
 // CHECK:      %[[T8:.*]] = vector.fma %[[T6]], %[[B]], %[[T7]] : vector<3xf32>
 // CHECK:      %[[T9:.*]] = vector.insert %[[T8]], %[[T4]] [1] : vector<3xf32> into vector<2x3xf32>
 // CHECK:      return %[[T9]] : vector<2x3xf32>
 
 func @outerproduct_acc(%arg0: vector<2xf32>,
                        %arg1: vector<3xf32>,
                        %arg2: vector<2x3xf32>) -> vector<2x3xf32> {
   %0 = vector.outerproduct %arg0, %arg1, %arg2 : vector<2xf32>, vector<3xf32>
   return %0: vector<2x3xf32>
 }
 
 // CHECK-LABEL: func @outerproduct_noacc_int
 // CHECK-SAME: %[[A:.*0]]: vector<2xi32>,
 // CHECK-SAME: %[[B:.*1]]: vector<3xi32>
 // CHECK:      %[[C0:.*]] = arith.constant dense<0> : vector<2x3xi32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : vector<2xi32>
-// CHECK:      %[[T1:.*]] = splat %[[T0]] : vector<3xi32>
+// CHECK:      %[[T1:.*]] = vector.splat %[[T0]] : vector<3xi32>
 // CHECK:      %[[T2:.*]] = arith.muli %[[T1]], %[[B]] : vector<3xi32>
 // CHECK:      %[[T3:.*]] = vector.insert %[[T2]], %[[C0]] [0] : vector<3xi32> into vector<2x3xi32>
 // CHECK:      %[[T4:.*]] = vector.extract %[[A]][1] : vector<2xi32>
-// CHECK:      %[[T5:.*]] = splat %[[T4]] : vector<3xi32>
+// CHECK:      %[[T5:.*]] = vector.splat %[[T4]] : vector<3xi32>
 // CHECK:      %[[T6:.*]] = arith.muli %[[T5]], %[[B]] : vector<3xi32>
 // CHECK:      %[[T7:.*]] = vector.insert %[[T6]], %[[T3]] [1] : vector<3xi32> into vector<2x3xi32>
 // CHECK:      return %[[T7]] : vector<2x3xi32>
 func @outerproduct_noacc_int(%arg0: vector<2xi32>,
                              %arg1: vector<3xi32>) -> vector<2x3xi32> {
   %0 = vector.outerproduct %arg0, %arg1 : vector<2xi32>, vector<3xi32>
   return %0: vector<2x3xi32>
 }
 
 // CHECK-LABEL: func @outerproduct_acc_int
 // CHECK-SAME: %[[A:.*0]]: vector<2xi32>,
 // CHECK-SAME: %[[B:.*1]]: vector<3xi32>,
 // CHECK-SAME: %[[C:.*2]]: vector<2x3xi32>
 // CHECK:      %[[C0:.*]] = arith.constant dense<0> : vector<2x3xi32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : vector<2xi32>
-// CHECK:      %[[T1:.*]] = splat %[[T0]] : vector<3xi32>
+// CHECK:      %[[T1:.*]] = vector.splat %[[T0]] : vector<3xi32>
 // CHECK:      %[[T2:.*]] = vector.extract %[[C]][0] : vector<2x3xi32>
 // CHECK:      %[[T3:.*]] = arith.muli %[[T1]], %[[B]] : vector<3xi32>
 // CHECK:      %[[T4:.*]] = arith.addi %[[T3]], %[[T2]] : vector<3xi32>
 // CHECK:      %[[T5:.*]] = vector.insert %[[T4]], %[[C0]] [0] : vector<3xi32> into vector<2x3xi32>
 // CHECK:      %[[T6:.*]] = vector.extract %[[A]][1] : vector<2xi32>
-// CHECK:      %[[T7:.*]] = splat %[[T6]] : vector<3xi32>
+// CHECK:      %[[T7:.*]] = vector.splat %[[T6]] : vector<3xi32>
 // CHECK:      %[[T8:.*]] = vector.extract %[[C]][1] : vector<2x3xi32>
 // CHECK:      %[[T9:.*]] = arith.muli %[[T7]], %[[B]] : vector<3xi32>
 // CHECK:      %[[T10:.*]] = arith.addi %[[T9]], %[[T8]] : vector<3xi32>
 // CHECK:      %[[T11:.*]] = vector.insert %[[T10]], %[[T5]] [1] : vector<3xi32> into vector<2x3xi32>
 // CHECK:      return %[[T11]] : vector<2x3xi32>
 func @outerproduct_acc_int(%arg0: vector<2xi32>,
                            %arg1: vector<3xi32>,
                            %arg2: vector<2x3xi32>) -> vector<2x3xi32> {
   %0 = vector.outerproduct %arg0, %arg1, %arg2 : vector<2xi32>, vector<3xi32>
   return %0: vector<2x3xi32>
 }
 
 // CHECK-LABEL: func @axpy_fp(
 // CHECK-SAME: %[[A:.*0]]: vector<16xf32>,
 // CHECK-SAME: %[[B:.*1]]: f32)
-// CHECK: %[[T0:.*]] = splat %[[B]] : vector<16xf32>
+// CHECK: %[[T0:.*]] = vector.splat %[[B]] : vector<16xf32>
 // CHECK: %[[T1:.*]] = arith.mulf %[[A]], %[[T0]] : vector<16xf32>
 // CHECK: return %[[T1]] : vector<16xf32>
 func @axpy_fp(%arg0: vector<16xf32>, %arg1: f32) -> vector<16xf32> {
    %0 = vector.outerproduct %arg0, %arg1: vector<16xf32>, f32
    return %0: vector<16xf32>
 }
 
 // CHECK-LABEL: func @axpy_fp_add(
 // CHECK-SAME: %[[A:.*0]]: vector<16xf32>,
 // CHECK-SAME: %[[B:.*1]]: f32,
 // CHECK-SAME: %[[C:.*2]]: vector<16xf32>)
-// CHECK: %[[T0:.*]] = splat %[[B]] : vector<16xf32>
+// CHECK: %[[T0:.*]] = vector.splat %[[B]] : vector<16xf32>
 // CHECK: %[[T1:.*]] = vector.fma %[[A]], %[[T0]], %[[C]] : vector<16xf32>
 // CHECK: return %[[T1]] : vector<16xf32>
 func @axpy_fp_add(%arg0: vector<16xf32>, %arg1: f32, %arg2 : vector<16xf32>) -> vector<16xf32> {
    %0 = vector.outerproduct %arg0, %arg1, %arg2: vector<16xf32>, f32
    return %0: vector<16xf32>
 }
 
 // CHECK-LABEL: func @axpy_int(
 // CHECK-SAME: %[[A:.*0]]: vector<16xi32>,
 // CHECK-SAME: %[[B:.*1]]: i32)
-// CHECK: %[[T0:.*]] = splat %[[B]] : vector<16xi32>
+// CHECK: %[[T0:.*]] = vector.splat %[[B]] : vector<16xi32>
 // CHECK: %[[T1:.*]] = arith.muli %[[A]], %[[T0]] : vector<16xi32>
 // CHECK: return %[[T1]] : vector<16xi32>
 func @axpy_int(%arg0: vector<16xi32>, %arg1: i32) -> vector<16xi32> {
    %0 = vector.outerproduct %arg0, %arg1: vector<16xi32>, i32
    return %0: vector<16xi32>
 }
 
 // CHECK-LABEL: func @axpy_int_add(
 // CHECK-SAME: %[[A:.*0]]: vector<16xi32>,
 // CHECK-SAME: %[[B:.*1]]: i32,
 // CHECK-SAME: %[[C:.*2]]: vector<16xi32>)
-// CHECK: %[[T0:.*]] = splat %[[B]] : vector<16xi32>
+// CHECK: %[[T0:.*]] = vector.splat %[[B]] : vector<16xi32>
 // CHECK: %[[T1:.*]] = arith.muli %[[A]], %[[T0]] : vector<16xi32>
 // CHECK: %[[T2:.*]] = arith.addi %[[T1]], %[[C]] : vector<16xi32>
 // CHECK: return %[[T2]] : vector<16xi32>
 func @axpy_int_add(%arg0: vector<16xi32>, %arg1: i32, %arg2: vector<16xi32>) -> vector<16xi32> {
    %0 = vector.outerproduct %arg0, %arg1, %arg2: vector<16xi32>, i32
    return %0: vector<16xi32>
 }
 
func @matmul(%arg0: vector<2x4xf32>,
                           %arg2: vector<2x3xf32>) -> vector<2x3xf32> {
   %0 = vector.contract #matmat_trait %arg0, %arg1, %arg2
     : vector<2x4xf32>, vector<4x3xf32> into vector<2x3xf32>
   return %0 : vector<2x3xf32>
 }
 
 // CHECK-LABEL: func @broadcast_vec1d_from_scalar
 // CHECK-SAME: %[[A:.*0]]: f32
-// CHECK:      %[[T0:.*]] = splat %[[A]] : vector<2xf32>
+// CHECK:      %[[T0:.*]] = vector.splat %[[A]] : vector<2xf32>
 // CHECK:      return %[[T0]] : vector<2xf32>
 
 func @broadcast_vec1d_from_scalar(%arg0: f32) -> vector<2xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2xf32>
   return %0 : vector<2xf32>
 }
 
 // CHECK-LABEL: func @broadcast_vec2d_from_scalar
 // CHECK-SAME: %[[A:.*0]]: f32
-// CHECK:      %[[T0:.*]] = splat %[[A]] : vector<2x3xf32>
+// CHECK:      %[[T0:.*]] = vector.splat %[[A]] : vector<2x3xf32>
 // CHECK:      return %[[T0]] : vector<2x3xf32>
 
 func @broadcast_vec2d_from_scalar(%arg0: f32) -> vector<2x3xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2x3xf32>
   return %0 : vector<2x3xf32>
 }
 
 // CHECK-LABEL: func @broadcast_vec3d_from_scalar
 // CHECK-SAME: %[[A:.*0]]: f32
-// CHECK:      %[[T0:.*]] = splat %[[A]] : vector<2x3x4xf32>
+// CHECK:      %[[T0:.*]] = vector.splat %[[A]] : vector<2x3x4xf32>
 // CHECK:      return %[[T0]] : vector<2x3x4xf32>
 
 func @broadcast_vec3d_from_scalar(%arg0: f32) -> vector<2x3x4xf32> {
   %0 = vector.broadcast %arg0 : f32 to vector<2x3x4xf32>
   return %0 : vector<2x3x4xf32>
 }
 
 // CHECK-LABEL: func @broadcast_vec1d_from_vec1d

 func @broadcast_vec3d_from_vec2d(%arg0: vector<3x2xf32>) -> vector<4x3x2xf32> {
   %0 = vector.broadcast %arg0 : vector<3x2xf32> to vector<4x3x2xf32>
   return %0 : vector<4x3x2xf32>
 }
 
 // CHECK-LABEL: func @broadcast_stretch
 // CHECK-SAME: %[[A:.*0]]: vector<1xf32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0] : vector<1xf32>
-// CHECK:      %[[T1:.*]] = splat %[[T0]] : vector<4xf32>
+// CHECK:      %[[T1:.*]] = vector.splat %[[T0]] : vector<4xf32>
 // CHECK:      return %[[T1]] : vector<4xf32>
 
 func @broadcast_stretch(%arg0: vector<1xf32>) -> vector<4xf32> {
   %0 = vector.broadcast %arg0 : vector<1xf32> to vector<4xf32>
   return %0 : vector<4xf32>
 }
 
 // CHECK-LABEL: func @broadcast_stretch_at_start
func @broadcast_stretch_at_start(%arg0: vector<1x4xf32>) -> vector<3x4xf32> {
   %0 = vector.broadcast %arg0 : vector<1x4xf32> to vector<3x4xf32>
   return %0 : vector<3x4xf32>
 }
 
 // CHECK-LABEL: func @broadcast_stretch_at_end
 // CHECK-SAME: %[[A:.*0]]: vector<4x1xf32>
 // CHECK:      %[[C0:.*]] = arith.constant dense<0.000000e+00> : vector<4x3xf32>
 // CHECK:      %[[T0:.*]] = vector.extract %[[A]][0, 0] : vector<4x1xf32>
-// CHECK:      %[[T2:.*]] = splat %[[T0]] : vector<3xf32>
+// CHECK:      %[[T2:.*]] = vector.splat %[[T0]] : vector<3xf32>
 // CHECK:      %[[T3:.*]] = vector.insert %[[T2]], %[[C0]] [0] : vector<3xf32> into vector<4x3xf32>
 // CHECK:      %[[T4:.*]] = vector.extract %[[A]][1, 0] : vector<4x1xf32>
-// CHECK:      %[[T6:.*]] = splat %[[T4]] : vector<3xf32>
+// CHECK:      %[[T6:.*]] = vector.splat %[[T4]] : vector<3xf32>
 // CHECK:      %[[T7:.*]] = vector.insert %[[T6]], %[[T3]] [1] : vector<3xf32> into vector<4x3xf32>
 // CHECK:      %[[T8:.*]] = vector.extract %[[A]][2, 0] : vector<4x1xf32>
-// CHECK:      %[[T10:.*]] = splat %[[T8]] : vector<3xf32>
+// CHECK:      %[[T10:.*]] = vector.splat %[[T8]] : vector<3xf32>
 // CHECK:      %[[T11:.*]] = vector.insert %[[T10]], %[[T7]] [2] : vector<3xf32> into vector<4x3xf32>
 // CHECK:      %[[T12:.*]] = vector.extract %[[A]][3, 0] : vector<4x1xf32>
-// CHECK:      %[[T14:.*]] = splat %[[T12]] : vector<3xf32>
+// CHECK:      %[[T14:.*]] = vector.splat %[[T12]] : vector<3xf32>
 // CHECK:      %[[T15:.*]] = vector.insert %[[T14]], %[[T11]] [3] : vector<3xf32> into vector<4x3xf32>
 // CHECK:      return %[[T15]] : vector<4x3xf32>
 
 func @broadcast_stretch_at_end(%arg0: vector<4x1xf32>) -> vector<4x3xf32> {
   %0 = vector.broadcast %arg0 : vector<4x1xf32> to vector<4x3xf32>
   return %0 : vector<4x3xf32>
 }
 

mlir/test/Dialect/Vector/vector-transfer-to-vector-load-store.mlir

func @transfer_read_permutations(%arg0 : memref<?x?xf32>, %arg1 : memref<?x?x?x?xf32>)
     -> (vector<7x14x8x16xf32>, vector<7x14x8x16xf32>, vector<7x14x8x16xf32>,
        vector<7x14x8x16xf32>, vector<7x14x8x16xf32>, vector<7x14x8x16xf32>, vector<8xf32>) {
 // CHECK-DAG: %[[CF0:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
   %cst = arith.constant 0.000000e+00 : f32
   %c0 = arith.constant 0 : index
   %m = arith.constant 1 : i1
 
-  %mask0 = splat %m : vector<7x14xi1>
+  %mask0 = vector.splat %m : vector<7x14xi1>
   %0 = vector.transfer_read %arg1[%c0, %c0, %c0, %c0], %cst, %mask0 {in_bounds = [true, false, true, true], permutation_map = #map0} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: %[[MASK0:.*]] = vector.transpose {{.*}} : vector<7x14xi1> to vector<14x7xi1>
 // CHECK: vector.transfer_read {{.*}} %[[MASK0]] {in_bounds = [false, true, true, true], permutation_map = #[[$MAP0]]} : memref<?x?x?x?xf32>, vector<14x7x8x16xf32>
 // CHECK: vector.transpose %{{.*}}, [1, 0, 2, 3] : vector<14x7x8x16xf32> to vector<7x14x8x16xf32>
 
-  %mask1 = splat %m : vector<14x16xi1>
+  %mask1 = vector.splat %m : vector<14x16xi1>
   %1 = vector.transfer_read %arg1[%c0, %c0, %c0, %c0], %cst, %mask1 {permutation_map = #map1} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: %[[MASK1:.*]] = vector.transpose {{.*}} : vector<14x16xi1> to vector<16x14xi1>
 // CHECK: vector.transfer_read {{.*}} %[[MASK1]] {permutation_map = #[[$MAP0]]} : memref<?x?x?x?xf32>, vector<16x14x7x8xf32>
 // CHECK: vector.transpose %{{.*}}, [2, 1, 3, 0] : vector<16x14x7x8xf32> to vector<7x14x8x16xf32>
 
-  %mask2 = splat %m : vector<7x14xi1>
+  %mask2 = vector.splat %m : vector<7x14xi1>
   %2 = vector.transfer_read %arg1[%c0, %c0, %c0, %c0], %cst, %mask2 {in_bounds = [true, false, true, true], permutation_map = #map2} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: %[[MASK2:.*]] = vector.transpose {{.*}} : vector<7x14xi1> to vector<14x7xi1>
 // CHECK: vector.transfer_read {{.*}} %[[MASK2]] {in_bounds = [false, true, true], permutation_map = #[[$MAP1]]} : memref<?x?x?x?xf32>, vector<14x16x7xf32>
 // CHECK: vector.broadcast %{{.*}} : vector<14x16x7xf32> to vector<8x14x16x7xf32>
 // CHECK: vector.transpose %{{.*}}, [3, 1, 0, 2] : vector<8x14x16x7xf32> to vector<7x14x8x16xf32>
 
   %3 = vector.transfer_read %arg0[%c0, %c0], %cst {permutation_map = #map3} : memref<?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: vector.transfer_read %{{.*}}[%[[C0]], %[[C0]]], %[[CF0]] : memref<?x?xf32>, vector<14x7xf32>

 // CHECK-SAME:      %[[ARG1:.*]]: tensor<?x?x?x?xf32>
 func @transfer_write_permutations(
     %arg0 : memref<?x?x?x?xf32>, %arg1 : tensor<?x?x?x?xf32>,
     %v1 : vector<7x14x8x16xf32>, %v2 : vector<8x16xf32>) -> tensor<?x?x?x?xf32> {
   // CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
   %c0 = arith.constant 0 : index
   %m = arith.constant 1 : i1
 
-  %mask0 = splat %m : vector<7x14x8x16xi1>
+  %mask0 = vector.splat %m : vector<7x14x8x16xi1>
   %0 = vector.transfer_write %v1, %arg1[%c0, %c0, %c0, %c0], %mask0 {in_bounds = [true, false, false, true], permutation_map = affine_map<(d0, d1, d2, d3) -> (d2, d1, d3, d0)>} : vector<7x14x8x16xf32>, tensor<?x?x?x?xf32>
   // CHECK: %[[NEW_MASK0:.*]] = vector.transpose %{{.*}} [2, 1, 3, 0] : vector<7x14x8x16xi1> to vector<8x14x16x7xi1>
   // CHECK: %[[NEW_VEC0:.*]] = vector.transpose %{{.*}} [2, 1, 3, 0] : vector<7x14x8x16xf32> to vector<8x14x16x7xf32>
   // CHECK: %[[NEW_RES0:.*]] = vector.transfer_write %[[NEW_VEC0]], %[[ARG1]][%c0, %c0, %c0, %c0], %[[NEW_MASK0]] {in_bounds = [false, false, true, true]} : vector<8x14x16x7xf32>, tensor<?x?x?x?xf32>
 
   vector.transfer_write %v2, %arg0[%c0, %c0, %c0, %c0] {permutation_map = affine_map<(d0, d1, d2, d3) -> (d3, d2)>} : vector<8x16xf32>, memref<?x?x?x?xf32>
   // CHECK: %[[NEW_VEC1:.*]] = vector.transpose %{{.*}} [1, 0] : vector<8x16xf32> to vector<16x8xf32>
   // CHECK: vector.transfer_write %[[NEW_VEC1]], %[[ARG0]][%c0, %c0, %c0, %c0] : vector<16x8xf32>, memref<?x?x?x?xf32>
 
   return %0 : tensor<?x?x?x?xf32>
 }

mlir/test/IR/core-ops.mlir

func @test_dimop(%arg0: tensor<4x4x?xf32>) {
   // CHECK: %{{.*}} = tensor.dim %[[ARG]], %[[C2]] : tensor<4x4x?xf32>
   %c2 = arith.constant 2 : index
   %0 = tensor.dim %arg0, %c2 : tensor<4x4x?xf32>
   // use dim as an index to ensure type correctness
   %1 = affine.apply affine_map<(d0) -> (d0)>(%0)
   return
 }
 
-// CHECK-LABEL: func @test_splat_op
-// CHECK-SAME: [[S:%arg[0-9]+]]: f32
-func @test_splat_op(%s : f32) {
-  %v = splat %s : vector<8xf32>
-  // CHECK: splat [[S]] : vector<8xf32>
-  %t = splat %s : tensor<8xf32>
-  // CHECK: splat [[S]] : tensor<8xf32>
-  %u = "std.splat"(%s) : (f32) -> vector<4xf32>
-  // CHECK: splat [[S]] : vector<4xf32>
-  return
-}
-
 // CHECK-LABEL: func @tensor_load_store
 func @tensor_load_store(%0 : memref<4x4xi32>, %1 : tensor<4x4xi32>) {
   // CHECK-SAME: (%[[MEMREF:.*]]: memref<4x4xi32>,
   // CHECK-SAME:  %[[TENSOR:.*]]: tensor<4x4xi32>)
   // CHECK: memref.tensor_store %[[TENSOR]], %[[MEMREF]] : memref<4x4xi32>
   memref.tensor_store %1, %0 : memref<4x4xi32>
   return
 }

mlir/test/IR/invalid-ops.mlir

"foo.region"() ({
     // expected-error@+1 {{'std.return' op expects parent op 'builtin.func'}}
     return
   }): () -> ()
   return
 }
 
 // -----
 
-func @invalid_splat(%v : f32) {
-  splat %v : memref<8xf32>
-  // expected-error@-1 {{must be vector of any type values or statically shaped tensor of any type values}}
-  return
-}
-
-// -----
-
-func @invalid_splat(%v : vector<8xf32>) {
-  %w = splat %v : tensor<8xvector<8xf32>>
-  // expected-error@-1 {{must be integer/index/float type}}
-  return
-}
-
-// -----
-
 func @invalid_splat(%v : f32) { // expected-note {{prior use here}}
-  splat %v : vector<8xf64>
+  vector.splat %v : vector<8xf64>
   // expected-error@-1 {{expects different type than prior uses}}
   return
 }

mlir/test/Integration/Dialect/Vector/CPU/test-0-d-vectors.mlir

 
 func @print_vector_0d(%a: vector<f32>) {
   // CHECK: ( 42 )
   vector.print %a: vector<f32>
   return
 }
 
 func @splat_0d(%a: f32) {
-  %1 = splat %a : vector<f32>
+  %1 = vector.splat %a : vector<f32>
   // CHECK: ( 42 )
   vector.print %1: vector<f32>
   return
 }
 
 func @broadcast_0d(%a: f32) {
   %1 = vector.broadcast %a : f32 to vector<f32>
   // CHECK: ( 42 )

mlir/test/Integration/Dialect/Vector/CPU/test-outerproduct-f32.mlir

 // RUN: mlir-opt %s -convert-scf-to-std -convert-vector-to-llvm -convert-std-to-llvm -reconcile-unrealized-casts | \
 // RUN: mlir-cpu-runner -e entry -entry-point-result=void  \
 // RUN:   -shared-libs=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
 // RUN: FileCheck %s
 
 !vector_type_A = type vector<8xf32>
 !vector_type_B = type vector<8xf32>
 !vector_type_C = type vector<8x8xf32>
 
 !vector_type_X = type vector<2xf32>
 !vector_type_Y = type vector<3xf32>
 !vector_type_Z = type vector<2x3xf32>
 
 !vector_type_R = type vector<7xf32>
 
 func @vector_outerproduct_splat_8x8(%fa: f32, %fb: f32, %fc: f32) -> !vector_type_C {
-  %a = splat %fa: !vector_type_A
-  %b = splat %fb: !vector_type_B
-  %c = splat %fc: !vector_type_C
+  %a = vector.splat %fa: !vector_type_A
+  %b = vector.splat %fb: !vector_type_B
+  %c = vector.splat %fc: !vector_type_C
   %d = vector.outerproduct %a, %b, %c : !vector_type_A, !vector_type_B
   return %d: !vector_type_C
 }
 
 func @vector_outerproduct_vec_2x3(%x : !vector_type_X,
                                   %y : !vector_type_Y) -> !vector_type_Z {
   %o = vector.outerproduct %x, %y : !vector_type_X, !vector_type_Y
   return %o: !vector_type_Z

mlir/test/Integration/Dialect/Vector/CPU/test-outerproduct-i64.mlir

 // RUN: mlir-opt %s -convert-scf-to-std -convert-vector-to-llvm -convert-std-to-llvm -reconcile-unrealized-casts | \
 // RUN: mlir-cpu-runner -e entry -entry-point-result=void  \
 // RUN:   -shared-libs=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
 // RUN: FileCheck %s
 
 !vector_type_A = type vector<8xi64>
 !vector_type_B = type vector<8xi64>
 !vector_type_C = type vector<8x8xi64>
 
 !vector_type_X = type vector<2xi64>
 !vector_type_Y = type vector<3xi64>
 !vector_type_Z = type vector<2x3xi64>
 
 !vector_type_R = type vector<7xi64>
 
 func @vector_outerproduct_splat_8x8(%ia: i64, %ib: i64, %ic: i64) -> !vector_type_C {
-  %a = splat %ia: !vector_type_A
-  %b = splat %ib: !vector_type_B
-  %c = splat %ic: !vector_type_C
+  %a = vector.splat %ia: !vector_type_A
+  %b = vector.splat %ib: !vector_type_B
+  %c = vector.splat %ic: !vector_type_C
   %d = vector.outerproduct %a, %b, %c : !vector_type_A, !vector_type_B
   return %d: !vector_type_C
 }
 
 func @vector_outerproduct_vec_2x3(%x : !vector_type_X,
                                   %y : !vector_type_Y) -> !vector_type_Z {
   %o = vector.outerproduct %x, %y : !vector_type_X, !vector_type_Y
   return %o: !vector_type_Z

mlir/test/Integration/Dialect/Vector/CPU/test-transfer-read-1d.mlir

%f = vector.transfer_read %A[%base1, %base2], %fm42, %mask
       : memref<?x?xf32>, vector<3xf32>
   vector.print %f: vector<3xf32>
   return
 }
 
 // Non-contiguous, strided store.
 func @transfer_write_1d(%A : memref<?x?xf32>, %base1 : index, %base2 : index) {
   %fn1 = arith.constant -1.0 : f32
-  %vf0 = splat %fn1 : vector<7xf32>
+  %vf0 = vector.splat %fn1 : vector<7xf32>
   vector.transfer_write %vf0, %A[%base1, %base2]
     {permutation_map = affine_map<(d0, d1) -> (d0)>}
     : vector<7xf32>, memref<?x?xf32>
   return
 }
 
 // Non-contiguous, strided store.
 func @transfer_write_1d_mask(%A : memref<?x?xf32>, %base1 : index, %base2 : index) {
   %fn1 = arith.constant -2.0 : f32
-  %vf0 = splat %fn1 : vector<7xf32>
+  %vf0 = vector.splat %fn1 : vector<7xf32>
   %mask = arith.constant dense<[1, 0, 1, 0, 1, 1, 1]> : vector<7xi1>
   vector.transfer_write %vf0, %A[%base1, %base2], %mask
     {permutation_map = affine_map<(d0, d1) -> (d0)>}
     : vector<7xf32>, memref<?x?xf32>
   return
 }
 
 func @entry() {

mlir/test/Integration/Dialect/Vector/CPU/test-transfer-read-2d.mlir

%f = vector.transfer_read %A[%base1, %base2], %fm42
     memref<?x?xf32>, vector<4x9xf32>
   vector.print %f: vector<4x9xf32>
   return
 }
 
 // Vector store.
 func @transfer_write_2d(%A : memref<?x?xf32>, %base1: index, %base2: index) {
   %fn1 = arith.constant -1.0 : f32
-  %vf0 = splat %fn1 : vector<1x4xf32>
+  %vf0 = vector.splat %fn1 : vector<1x4xf32>
   vector.transfer_write %vf0, %A[%base1, %base2]
     {permutation_map = affine_map<(d0, d1) -> (d0, d1)>} :
     vector<1x4xf32>, memref<?x?xf32>
   return
 }
 
 // Vector store with mask.
 func @transfer_write_2d_mask(%A : memref<?x?xf32>, %base1: index, %base2: index) {
   %fn1 = arith.constant -2.0 : f32
   %mask = arith.constant dense<[[1, 0, 1, 0]]> : vector<1x4xi1>
-  %vf0 = splat %fn1 : vector<1x4xf32>
+  %vf0 = vector.splat %fn1 : vector<1x4xf32>
   vector.transfer_write %vf0, %A[%base1, %base2], %mask
     {permutation_map = affine_map<(d0, d1) -> (d0, d1)>} :
     vector<1x4xf32>, memref<?x?xf32>
   return
 }
 
 func @entry() {
   %c0 = arith.constant 0: index

mlir/test/Integration/Dialect/Vector/CPU/test-transfer-read-3d.mlir

%f = vector.transfer_read %A[%o, %a, %b, %c], %fm42
       : memref<?x?x?x?xf32>, vector<3x5x3xf32>
   vector.print %f: vector<3x5x3xf32>
   return
 }
 
 func @transfer_write_3d(%A : memref<?x?x?x?xf32>,
                         %o: index, %a: index, %b: index, %c: index) {
   %fn1 = arith.constant -1.0 : f32
-  %vf0 = splat %fn1 : vector<2x9x3xf32>
+  %vf0 = vector.splat %fn1 : vector<2x9x3xf32>
   vector.transfer_write %vf0, %A[%o, %a, %b, %c]
       : vector<2x9x3xf32>, memref<?x?x?x?xf32>
   return
 }
 
 func @entry() {
   %c0 = arith.constant 0: index
   %c1 = arith.constant 1: index

mlir/test/Integration/Dialect/Vector/CPU/test-transfer-read.mlir

func @transfer_read_mask_inbounds_4(%A : memref<?xf32>, %base: index) {
   %f = vector.transfer_read %A[%base], %fm42, %m {in_bounds = [true]}
       : memref<?xf32>, vector<4xf32>
   vector.print %f: vector<4xf32>
   return
 }
 
 func @transfer_write_1d(%A : memref<?xf32>, %base: index) {
   %f0 = arith.constant 0.0 : f32
-  %vf0 = splat %f0 : vector<4xf32>
+  %vf0 = vector.splat %f0 : vector<4xf32>
   vector.transfer_write %vf0, %A[%base]
       {permutation_map = affine_map<(d0) -> (d0)>} :
     vector<4xf32>, memref<?xf32>
   return
 }
 
 func @entry() {
   %c0 = arith.constant 0: index

mlir/test/Integration/Dialect/Vector/CPU/test-transfer-write.mlir

 // RUN: mlir-opt %s -convert-scf-to-std -convert-vector-to-llvm -convert-memref-to-llvm -convert-std-to-llvm -reconcile-unrealized-casts | \
 // RUN: mlir-cpu-runner -e entry -entry-point-result=void  \
 // RUN:   -shared-libs=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
 // RUN: FileCheck %s
 
 func @transfer_write16_inbounds_1d(%A : memref<?xf32>, %base: index) {
   %f = arith.constant 16.0 : f32
-  %v = splat %f : vector<16xf32>
+  %v = vector.splat %f : vector<16xf32>
   vector.transfer_write %v, %A[%base]
     {permutation_map = affine_map<(d0) -> (d0)>, in_bounds = [true]}
     : vector<16xf32>, memref<?xf32>
   return
 }
 
 func @transfer_write13_1d(%A : memref<?xf32>, %base: index) {
   %f = arith.constant 13.0 : f32
-  %v = splat %f : vector<13xf32>
+  %v = vector.splat %f : vector<13xf32>
   vector.transfer_write %v, %A[%base]
     {permutation_map = affine_map<(d0) -> (d0)>}
     : vector<13xf32>, memref<?xf32>
   return
 }
 
 func @transfer_write17_1d(%A : memref<?xf32>, %base: index) {
   %f = arith.constant 17.0 : f32
-  %v = splat %f : vector<17xf32>
+  %v = vector.splat %f : vector<17xf32>
   vector.transfer_write %v, %A[%base]
     {permutation_map = affine_map<(d0) -> (d0)>}
     : vector<17xf32>, memref<?xf32>
   return
 }
 
 func @transfer_read_1d(%A : memref<?xf32>) -> vector<32xf32> {
   %z = arith.constant 0: index

mlir/test/Transforms/constant-fold.mlir

"test.one_region_op"() ({
 
     %0 = arith.constant 1 : i32
     %2 = arith.addi %0, %0 : i32
     "foo.yield"(%2) : (i32) -> ()
   }) : () -> ()
   return
 }
 
-// CHECK-LABEL: func @splat_fold
-func @splat_fold() -> (vector<4xf32>, tensor<4xf32>) {
-  %c = arith.constant 1.0 : f32
-  %v = splat %c : vector<4xf32>
-  %t = splat %c : tensor<4xf32>
-  return %v, %t : vector<4xf32>, tensor<4xf32>
-
-  // CHECK-NEXT: [[V:%.*]] = arith.constant dense<1.000000e+00> : vector<4xf32>
-  // CHECK-NEXT: [[T:%.*]] = arith.constant dense<1.000000e+00> : tensor<4xf32>
-  // CHECK-NEXT: return [[V]], [[T]] : vector<4xf32>, tensor<4xf32>
-}
-
 // -----
 
 // CHECK-LABEL: func @subview_scalar_fold
 func @subview_scalar_fold(%arg0: memref<f32>) -> memref<f32> {
   // CHECK-NOT: memref.subview
   %c = memref.subview %arg0[] [] [] : memref<f32> to memref<f32>
   return %c : memref<f32>
 }

mlir/test/mlir-cpu-runner/utils.mlir

 
 func private @print_memref_f32(memref<*xf32>) attributes { llvm.emit_c_interface }
 
 !vector_type_C = type vector<4x4xf32>
 !matrix_type_CC = type memref<1x1x!vector_type_C>
 func @vector_splat_2d() {
   %c0 = arith.constant 0 : index
   %f10 = arith.constant 10.0 : f32
-  %vf10 = splat %f10: !vector_type_C
+  %vf10 = vector.splat %f10: !vector_type_C
   %C = memref.alloc() : !matrix_type_CC
   memref.store %vf10, %C[%c0, %c0]: !matrix_type_CC
 
   %CC = memref.cast %C: !matrix_type_CC to memref<?x?x!vector_type_C>
   call @print_memref_vector_4x4xf32(%CC): (memref<?x?x!vector_type_C>) -> ()
 
   memref.dealloc %C : !matrix_type_CC
   return
 }
 
 // PRINT-VECTOR-SPLAT-2D: Memref base@ = {{.*}} rank = 2 offset = 0 sizes = [1, 1] strides = [1, 1] data =
 // PRINT-VECTOR-SPLAT-2D-NEXT: [((10, 10, 10, 10),   (10, 10, 10, 10),   (10, 10, 10, 10),   (10, 10, 10, 10))]
 
 func private @print_memref_vector_4x4xf32(memref<?x?x!vector_type_C>) attributes { llvm.emit_c_interface }