Add UnrankedMemRef Type

Closes tensorflow/mlir#261

COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/261 from nmostafa:nmostafa/unranked 96b6e918f6ed64496f7573b2db33c0b02658ca45
PiperOrigin-RevId: 284037040

mlir/g3doc/ConversionToLLVMDialect.md

@@ -90,6 +90,23 @@ memref<10x?x42x?x123 x f32> -> !llvm.type<"{ float*, float*, i64, [5 x i64], [5
 memref<1x? x vector<4xf32>> -> !llvm.type<"{ <4 x float>*, <4 x float>*, i64, [1 x i64], [1 x i64] }">
 ```
 
+If the rank of the memref is unknown at compile time, the Memref is converted to
+an unranked descriptor that contains: 1. a 64-bit integer representing the
+dynamic rank of the memref, followed by 2. a pointer to a ranked memref
+descriptor with the contents listed above.
+
+Dynamic ranked memrefs should be used only to pass arguments to external library
+calls that expect a unified memref type. The called functions can parse any
+unranked memref descriptor by reading the rank and parsing the enclosed ranked
+descriptor pointer.
+
+Examples:
+
+```mlir {.mlir}
+// unranked descriptor
+memref<*xf32> -> !llvm.type<"{i64, i8*}">
+```
+
 ### Function Types
 
 Function types get converted to LLVM function types. The arguments are converted

mlir/g3doc/Dialects/Standard.md

@@ -912,12 +912,21 @@ Examples:
 
 // Convert to a type with more known dimensions.
 %4 = memref_cast %3 : memref<?x?xf32> to memref<4x?xf32>
+
+// Convert to a type with unknown rank.
+%5 = memref_cast %3 : memref<?x?xf32> to memref<*xf32>
+
+// Convert to a type with static rank.
+%6 = memref_cast %5 : memref<*xf32> to memref<?x?xf32>
 ```
 
 Convert a memref from one type to an equivalent type without changing any data
-elements. The source and destination types must both be memref types with the
+elements. The types are equivalent if 1. they both have the same static rank,
-same element type, same mappings, same address space, and same rank. The
+same element type, same mappings, same address space. The operation is invalid
-operation is invalid if converting to a mismatching constant dimension.
+if converting to a mismatching constant dimension, or 2. exactly one of the
+operands have an unknown rank, and they both have the same element type and same
+address space. The operation is invalid if both operands are of dynamic rank or
+if converting to a mismatching static rank.
 
 ### 'mulf' operation
 

mlir/g3doc/LangRef.md

@@ -760,9 +760,15 @@ TODO: Need to decide on a representation for quantized integers
 Syntax:
 
 ``` {.ebnf}
-memref-type ::= `memref` `<` dimension-list-ranked tensor-memref-element-type
+
-                (`,` layout-specification)? |
+memref-type ::= ranked-memref-type | unranked-memref-type
-                (`,` memory-space)? `>`
+
+ranked-memref-type ::= `memref` `<` dimension-list-ranked tensor-memref-element-type
+                      (`,` layout-specification)? |
+                      (`,` memory-space)? `>`
+
+unranked-memref-type ::= `memref` `<*x` tensor-memref-element-type
+                         (`,` memory-space)? `>`
 
 stride-list ::= `[` (dimension (`,` dimension)*)? `]`
 strided-layout ::= `offset:` dimension `,` `strides: ` stride-list
@@ -774,9 +780,48 @@ A `memref` type is a reference to a region of memory (similar to a buffer
 pointer, but more powerful). The buffer pointed to by a memref can be allocated,
 aliased and deallocated. A memref can be used to read and write data from/to the
 memory region which it references. Memref types use the same shape specifier as
-tensor types, but do not allow unknown rank. Note that `memref<f32>`, `memref<0
+tensor types. Note that `memref<f32>`, `memref<0 x f32>`, `memref<1 x 0 x f32>`,
-x f32>`, `memref<1 x 0 x f32>`, and `memref<0 x 1 x f32>` are all different
+and `memref<0 x 1 x f32>` are all different types.
-types.
+
+A `memref` is allowed to have an unknown rank (e.g. `memref<*xf32>`). The
+purpose of unranked memrefs is to allow external library functions to receive
+memref arguments of any rank without versioning the functions based on the rank.
+Other uses of this type are disallowed or will have undefined behavior.
+
+##### Codegen of Unranked Memref
+
+Using unranked memref in codegen besides the case mentioned above is highly
+discouraged. Codegen is concerned with generating loop nests and specialized
+instructions for high-performance, unranked memref is concerned with hiding the
+rank and thus, the number of enclosing loops required to iterate over the data.
+However, if there is a need to code-gen unranked memref, one possible path is to
+cast into a static ranked type based on the dynamic rank. Another possible path
+is to emit a single while loop conditioned on a linear index and perform
+delinearization of the linear index to a dynamic array containing the (unranked)
+indices. While this is possible, it is expected to not be a good idea to perform
+this during codegen as the cost of the translations is expected to be
+prohibitive and optimizations at this level are not expected to be worthwhile.
+If expressiveness is the main concern, irrespective of performance, passing
+unranked memrefs to an external C++ library and implementing rank-agnostic logic
+there is expected to be significantly simpler.
+
+Unranked memrefs may provide expressiveness gains in the future and help bridge
+the gap with unranked tensors. Unranked memrefs will not be expected to be
+exposed to codegen but one may query the rank of an unranked memref (a special
+op will be needed for this purpose) and perform a switch and cast to a ranked
+memref as a prerequisite to codegen.
+
+Example ```mlir {.mlir} // With static ranks, we need a function for each
+possible argument type %A = alloc() : memref<16x32xf32> %B = alloc() :
+memref<16x32x64xf32> call @helper_2D(%A) : (memref<16x32xf32>)->() call
+@helper_3D(%B) : (memref<16x32x64xf32>)->()
+
+// With unknown rank, the functions can be unified under one unranked type %A =
+alloc() : memref<16x32xf32> %B = alloc() : memref<16x32x64xf32> // Remove rank
+info %A_u = memref_cast %A : memref<16x32xf32> -> memref<*xf32> %B_u =
+memref_cast %B : memref<16x32x64xf32> -> memref<*xf32> // call same function
+with dynamic ranks call @helper(%A_u) : (memref<*xf32>)->() call @helper(%B_u) :
+(memref<*xf32>)->() ```
 
 The core syntax and representation of a layout specification is a
 [semi-affine map](Dialects/Affine.md#semi-affine-maps). Additionally, syntactic

mlir/include/mlir/Conversion/StandardToLLVM/ConvertStandardToLLVM.h

@@ -34,6 +34,9 @@ class Type;
 } // namespace llvm
 
 namespace mlir {
+
+class UnrankedMemRefType;
+
 namespace LLVM {
 class LLVMDialect;
 class LLVMType;
@@ -116,6 +119,10 @@ private:
   //   2. as many index types as memref has dynamic dimensions.
   Type convertMemRefType(MemRefType type);
 
+  // Convert an unranked memref type to an LLVM type that captures the
+  // runtime rank and a pointer to the static ranked memref desc
+  Type convertUnrankedMemRefType(UnrankedMemRefType type);
+
   // Convert a 1D vector type into an LLVM vector type.
   Type convertVectorType(VectorType type);
 
@@ -127,10 +134,34 @@ private:
   LLVM::LLVMType unwrap(Type type);
 };
 
+/// Helper class to produce LLVM dialect operations extracting or inserting
+/// values to a struct.
+class StructBuilder {
+public:
+  /// Construct a helper for the given value.
+  explicit StructBuilder(Value *v);
+  /// Builds IR creating an `undef` value of the descriptor type.
+  static StructBuilder undef(OpBuilder &builder, Location loc,
+                             Type descriptorType);
+
+  /*implicit*/ operator Value *() { return value; }
+
+protected:
+  // LLVM value
+  Value *value;
+  // Cached struct type.
+  Type structType;
+
+protected:
+  /// Builds IR to extract a value from the struct at position pos
+  Value *extractPtr(OpBuilder &builder, Location loc, unsigned pos);
+  /// Builds IR to set a value in the struct at position pos
+  void setPtr(OpBuilder &builder, Location loc, unsigned pos, Value *ptr);
+};
 /// Helper class to produce LLVM dialect operations extracting or inserting
 /// elements of a MemRef descriptor. Wraps a Value pointing to the descriptor.
 /// The Value may be null, in which case none of the operations are valid.
-class MemRefDescriptor {
+class MemRefDescriptor : public StructBuilder {
 public:
   /// Construct a helper for the given descriptor value.
   explicit MemRefDescriptor(Value *descriptor);
@@ -169,22 +200,28 @@ public:
   /// Returns the (LLVM) type this descriptor points to.
   LLVM::LLVMType getElementType();
 
-  /*implicit*/ operator Value *() { return value; }
-
 private:
-  Value *extractPtr(OpBuilder &builder, Location loc, unsigned pos);
-  void setPtr(OpBuilder &builder, Location loc, unsigned pos, Value *ptr);
-
-  // Cached descriptor type.
-  Type structType;
-
   // Cached index type.
   Type indexType;
-
-  // Actual descriptor.
-  Value *value;
 };
 
+class UnrankedMemRefDescriptor : public StructBuilder {
+public:
+  /// Construct a helper for the given descriptor value.
+  explicit UnrankedMemRefDescriptor(Value *descriptor);
+  /// Builds IR creating an `undef` value of the descriptor type.
+  static UnrankedMemRefDescriptor undef(OpBuilder &builder, Location loc,
+                                        Type descriptorType);
+
+  /// Builds IR extracting the rank from the descriptor
+  Value *rank(OpBuilder &builder, Location loc);
+  /// Builds IR setting the rank in the descriptor
+  void setRank(OpBuilder &builder, Location loc, Value *value);
+  /// Builds IR extracting ranked memref descriptor ptr
+  Value *memRefDescPtr(OpBuilder &builder, Location loc);
+  /// Builds IR setting ranked memref descriptor ptr
+  void setMemRefDescPtr(OpBuilder &builder, Location loc, Value *value);
+};
 /// Base class for operation conversions targeting the LLVM IR dialect. Provides
 /// conversion patterns with an access to the containing LLVMLowering for the
 /// purpose of type conversions.

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

@@ -842,7 +842,8 @@ def MemRefCastOp : CastOp<"memref_cast"> {
   let description = [{
     The "memref_cast" operation converts a memref from one type to an equivalent
     type with a compatible shape. The source and destination types are
-    when both are memref types with the same element type, affine mappings,
+    compatible if:
+    a. both are ranked memref types with the same element type, affine mappings,
     address space, and rank but where the individual dimensions may add or
     remove constant dimensions from the memref type.
 
@@ -850,6 +851,7 @@ def MemRefCastOp : CastOp<"memref_cast"> {
     acts as an assertion that fails at runtime of the dynamic dimensions
     disagree with resultant destination size.
 
+    Example:
     Assert that the input dynamic shape matches the destination static shape.
        %2 = memref_cast %1 : memref<?x?xf32> to memref<4x4xf32>
     Erase static shape information, replacing it with dynamic information.
@@ -864,10 +866,20 @@ def MemRefCastOp : CastOp<"memref_cast"> {
     dynamic information.
        %5 = memref_cast %1 : memref<12x4xf32, offset:5, strides: [4, 1]> to
                              memref<12x4xf32, offset:?, strides: [?, ?]>
+
+    b. either or both memref types are unranked with the same element type, and
+    address space.
+
+    Example:
+    Cast to concrete shape.
+        %4 = memref_cast %1 : memref<*xf32> to memref<4x?xf32>
+
+    Erase rank information.
+        %5 = memref_cast %1 : memref<4x?xf32> to memref<*xf32>
   }];
 
-  let arguments = (ins AnyMemRef:$source);
+  let arguments = (ins AnyRankedOrUnrankedMemRef:$source);
-  let results = (outs AnyMemRef);
+  let results = (outs AnyRankedOrUnrankedMemRef);
 
   let extraClassDeclaration = [{
     /// Return true if `a` and `b` are valid operand and result pairs for
@@ -875,7 +887,7 @@ def MemRefCastOp : CastOp<"memref_cast"> {
     static bool areCastCompatible(Type a, Type b);
 
     /// The result of a memref_cast is always a memref.
-    MemRefType getType() { return getResult()->getType().cast<MemRefType>(); }
+    Type getType() { return getResult()->getType(); }
   }];
 }
 

mlir/include/mlir/IR/OpBase.td

@@ -221,6 +221,9 @@ def IsTensorTypePred : CPred<"$_self.isa<TensorType>()">;
 // Whether a type is a MemRefType.
 def IsMemRefTypePred : CPred<"$_self.isa<MemRefType>()">;
 
+// Whether a type is an  IsUnrankedMemRefType
+def IsUnrankedMemRefTypePred : CPred<"$_self.isa<UnrankedMemRefType>()">;
+
 // Whether a type is a ShapedType.
 def IsShapedTypePred : CPred<"$_self.isa<ShapedType>()">;
 
@@ -486,6 +489,10 @@ class 2DTensorOf<list<Type> allowedTypes> : TensorRankOf<allowedTypes, [2]>;
 class 3DTensorOf<list<Type> allowedTypes> : TensorRankOf<allowedTypes, [3]>;
 class 4DTensorOf<list<Type> allowedTypes> : TensorRankOf<allowedTypes, [4]>;
 
+// Unranked Memref type
+def AnyUnrankedMemRef : 
+    ShapedContainerType<[AnyType], 
+                        IsUnrankedMemRefTypePred, "unranked.memref">;
 // Memref type.
 
 // Memrefs are blocks of data with fixed type and rank.
@@ -494,6 +501,8 @@ class MemRefOf<list<Type> allowedTypes> :
 
 def AnyMemRef : MemRefOf<[AnyType]>;
 
+def AnyRankedOrUnrankedMemRef: AnyTypeOf<[AnyUnrankedMemRef, AnyMemRef]>;
+
 // Memref declarations handle any memref, independent of rank, size, (static or
 // dynamic), layout, or memory space.
 def I1MemRef  : MemRefOf<[I1]>;

mlir/include/mlir/IR/StandardTypes.h

@@ -40,6 +40,7 @@ struct VectorTypeStorage;
 struct RankedTensorTypeStorage;
 struct UnrankedTensorTypeStorage;
 struct MemRefTypeStorage;
+struct UnrankedMemRefTypeStorage;
 struct ComplexTypeStorage;
 struct TupleTypeStorage;
 
@@ -64,6 +65,7 @@ enum Kind {
   RankedTensor,
   UnrankedTensor,
   MemRef,
+  UnrankedMemRef,
   Complex,
   Tuple,
   None,
@@ -243,6 +245,7 @@ public:
     return type.getKind() == StandardTypes::Vector ||
            type.getKind() == StandardTypes::RankedTensor ||
            type.getKind() == StandardTypes::UnrankedTensor ||
+           type.getKind() == StandardTypes::UnrankedMemRef ||
            type.getKind() == StandardTypes::MemRef;
   }
 
@@ -370,12 +373,24 @@ public:
   }
 };
 
+/// Base MemRef for Ranked and Unranked variants
+class BaseMemRefType : public ShapedType {
+public:
+  using ShapedType::ShapedType;
+
+  /// Methods for support type inquiry through isa, cast, and dyn_cast.
+  static bool classof(Type type) {
+    return type.getKind() == StandardTypes::MemRef ||
+           type.getKind() == StandardTypes::UnrankedMemRef;
+  }
+};
+
 /// MemRef types represent a region of memory that have a shape with a fixed
 /// number of dimensions. Each shape element can be a non-negative integer or
 /// unknown (represented by any negative integer). MemRef types also have an
 /// affine map composition, represented as an array AffineMap pointers.
-class MemRefType
+class MemRefType : public Type::TypeBase<MemRefType, BaseMemRefType,
-    : public Type::TypeBase<MemRefType, ShapedType, detail::MemRefTypeStorage> {
+                                         detail::MemRefTypeStorage> {
 public:
   using Base::Base;
 
@@ -426,6 +441,40 @@ private:
   using Base::getImpl;
 };
 
+/// Unranked MemRef type represent multi-dimensional MemRefs that
+/// have an unknown rank.
+class UnrankedMemRefType
+    : public Type::TypeBase<UnrankedMemRefType, BaseMemRefType,
+                            detail::UnrankedMemRefTypeStorage> {
+public:
+  using Base::Base;
+
+  /// Get or create a new UnrankedMemRefType of the provided element
+  /// type and memory space
+  static UnrankedMemRefType get(Type elementType, unsigned memorySpace);
+
+  /// Get or create a new UnrankedMemRefType of the provided element
+  /// type and memory space declared at the given, potentially unknown,
+  /// location. If the UnrankedMemRefType defined by the arguments would be
+  /// ill-formed, emit errors and return a nullptr-wrapping type.
+  static UnrankedMemRefType getChecked(Type elementType, unsigned memorySpace,
+                                       Location location);
+
+  /// Verify the construction of a unranked memref type.
+  static LogicalResult
+  verifyConstructionInvariants(llvm::Optional<Location> loc,
+                               MLIRContext *context, Type elementType,
+                               unsigned memorySpace);
+
+  ArrayRef<int64_t> getShape() const { return llvm::None; }
+
+  /// Returns the memory space in which data referred to by this memref resides.
+  unsigned getMemorySpace() const;
+  static bool kindof(unsigned kind) {
+    return kind == StandardTypes::UnrankedMemRef;
+  }
+};
+
 /// Tuple types represent a collection of other types. Note: This type merely
 /// provides a common mechanism for representing tuples in MLIR. It is up to
 /// dialect authors to provides operations for manipulating them, e.g.

mlir/lib/Conversion/StandardToLLVM/ConvertStandardToLLVM.cpp

@@ -193,6 +193,22 @@ Type LLVMTypeConverter::convertMemRefType(MemRefType type) {
   return LLVM::LLVMType::getStructTy(ptrTy, ptrTy, indexTy);
 }
 
+// Converts UnrankedMemRefType to LLVMType. The result is a descriptor which
+// contains:
+// 1. int64_t rank, the dynamic rank of this MemRef
+// 2. void* ptr, pointer to the static ranked MemRef descriptor. This will be
+//    stack allocated (alloca) copy of a MemRef descriptor that got casted to
+//    be unranked.
+
+static constexpr unsigned kRankInUnrankedMemRefDescriptor = 0;
+static constexpr unsigned kPtrInUnrankedMemRefDescriptor = 1;
+
+Type LLVMTypeConverter::convertUnrankedMemRefType(UnrankedMemRefType type) {
+  auto rankTy = LLVM::LLVMType::getInt64Ty(llvmDialect);
+  auto ptrTy = LLVM::LLVMType::getInt8PtrTy(llvmDialect);
+  return LLVM::LLVMType::getStructTy(rankTy, ptrTy);
+}
+
 // Convert an n-D vector type to an LLVM vector type via (n-1)-D array type when
 // n > 1.
 // For example, `vector<4 x f32>` converts to `!llvm.type<"<4 x float>">` and
@@ -221,6 +237,8 @@ Type LLVMTypeConverter::convertStandardType(Type type) {
     return convertIndexType(indexType);
   if (auto memRefType = type.dyn_cast<MemRefType>())
     return convertMemRefType(memRefType);
+  if (auto memRefType = type.dyn_cast<UnrankedMemRefType>())
+    return convertUnrankedMemRefType(memRefType);
   if (auto vectorType = type.dyn_cast<VectorType>())
     return convertVectorType(vectorType);
   if (auto llvmType = type.dyn_cast<LLVM::LLVMType>())
@@ -245,22 +263,42 @@ LLVMOpLowering::LLVMOpLowering(StringRef rootOpName, MLIRContext *context,
                                PatternBenefit benefit)
     : ConversionPattern(rootOpName, benefit, context), lowering(lowering_) {}
 
+/*============================================================================*/
+/* StructBuilder implementation                                               */
+/*============================================================================*/
+StructBuilder::StructBuilder(Value *v) : value(v) {
+  assert(value != nullptr && "value cannot be null");
+  structType = value->getType().cast<LLVM::LLVMType>();
+}
+
+Value *StructBuilder::extractPtr(OpBuilder &builder, Location loc,
+                                 unsigned pos) {
+  Type type = structType.cast<LLVM::LLVMType>().getStructElementType(pos);
+  return builder.create<LLVM::ExtractValueOp>(loc, type, value,
+                                              builder.getI64ArrayAttr(pos));
+}
+
+void StructBuilder::setPtr(OpBuilder &builder, Location loc, unsigned pos,
+                           Value *ptr) {
+  value = builder.create<LLVM::InsertValueOp>(loc, structType, value, ptr,
+                                              builder.getI64ArrayAttr(pos));
+}
 /*============================================================================*/
 /* MemRefDescriptor implementation                                            */
 /*============================================================================*/
 
 /// Construct a helper for the given descriptor value.
-MemRefDescriptor::MemRefDescriptor(Value *descriptor) : value(descriptor) {
+MemRefDescriptor::MemRefDescriptor(Value *descriptor)
-  if (value) {
+    : StructBuilder(descriptor) {
-    structType = value->getType().cast<LLVM::LLVMType>();
+  assert(value != nullptr && "value cannot be null");
-    indexType = value->getType().cast<LLVM::LLVMType>().getStructElementType(
+  indexType = value->getType().cast<LLVM::LLVMType>().getStructElementType(
-        kOffsetPosInMemRefDescriptor);
+      kOffsetPosInMemRefDescriptor);
-  }
 }
 
 /// Builds IR creating an `undef` value of the descriptor type.
 MemRefDescriptor MemRefDescriptor::undef(OpBuilder &builder, Location loc,
                                          Type descriptorType) {
+
   Value *descriptor =
       builder.create<LLVM::UndefOp>(loc, descriptorType.cast<LLVM::LLVMType>());
   return MemRefDescriptor(descriptor);
@@ -334,24 +372,42 @@ void MemRefDescriptor::setStride(OpBuilder &builder, Location loc, unsigned pos,
       builder.getI64ArrayAttr({kStridePosInMemRefDescriptor, pos}));
 }
 
-Value *MemRefDescriptor::extractPtr(OpBuilder &builder, Location loc,
-                                    unsigned pos) {
-  Type type = structType.cast<LLVM::LLVMType>().getStructElementType(pos);
-  return builder.create<LLVM::ExtractValueOp>(loc, type, value,
-                                              builder.getI64ArrayAttr(pos));
-}
-
-void MemRefDescriptor::setPtr(OpBuilder &builder, Location loc, unsigned pos,
-                              Value *ptr) {
-  value = builder.create<LLVM::InsertValueOp>(loc, structType, value, ptr,
-                                              builder.getI64ArrayAttr(pos));
-}
-
 LLVM::LLVMType MemRefDescriptor::getElementType() {
   return value->getType().cast<LLVM::LLVMType>().getStructElementType(
       kAlignedPtrPosInMemRefDescriptor);
 }
 
+/*============================================================================*/
+/* UnrankedMemRefDescriptor implementation                                    */
+/*============================================================================*/
+
+/// Construct a helper for the given descriptor value.
+UnrankedMemRefDescriptor::UnrankedMemRefDescriptor(Value *descriptor)
+    : StructBuilder(descriptor) {}
+
+/// Builds IR creating an `undef` value of the descriptor type.
+UnrankedMemRefDescriptor UnrankedMemRefDescriptor::undef(OpBuilder &builder,
+                                                         Location loc,
+                                                         Type descriptorType) {
+  Value *descriptor =
+      builder.create<LLVM::UndefOp>(loc, descriptorType.cast<LLVM::LLVMType>());
+  return UnrankedMemRefDescriptor(descriptor);
+}
+Value *UnrankedMemRefDescriptor::rank(OpBuilder &builder, Location loc) {
+  return extractPtr(builder, loc, kRankInUnrankedMemRefDescriptor);
+}
+void UnrankedMemRefDescriptor::setRank(OpBuilder &builder, Location loc,
+                                       Value *v) {
+  setPtr(builder, loc, kRankInUnrankedMemRefDescriptor, v);
+}
+Value *UnrankedMemRefDescriptor::memRefDescPtr(OpBuilder &builder,
+                                               Location loc) {
+  return extractPtr(builder, loc, kPtrInUnrankedMemRefDescriptor);
+}
+void UnrankedMemRefDescriptor::setMemRefDescPtr(OpBuilder &builder,
+                                                Location loc, Value *v) {
+  setPtr(builder, loc, kPtrInUnrankedMemRefDescriptor, v);
+}
 namespace {
 // Base class for Standard to LLVM IR op conversions.  Matches the Op type
 // provided as template argument.  Carries a reference to the LLVM dialect in
@@ -432,7 +488,7 @@ struct FuncOpConversion : public LLVMLegalizationPattern<FuncOp> {
       auto converted = lowering.convertType(t).dyn_cast<LLVM::LLVMType>();
       if (!converted)
         return matchFailure();
-      if (t.isa<MemRefType>()) {
+      if (t.isa<MemRefType>() || t.isa<UnrankedMemRefType>()) {
         converted = converted.getPointerTo();
         promotedArgIndices.push_back(en.index());
       }
@@ -983,6 +1039,14 @@ struct CallOpInterfaceLowering : public LLVMLegalizationPattern<CallOpType> {
     Type packedResult;
     unsigned numResults = callOp.getNumResults();
     auto resultTypes = llvm::to_vector<4>(callOp.getResultTypes());
+
+    for (Type resType : resultTypes) {
+      assert(!resType.isa<UnrankedMemRefType>() &&
+             "Returning unranked memref is not supported. Pass result as an"
+             "argument instead.");
+      (void)resType;
+    }
+
     if (numResults != 0) {
       if (!(packedResult = this->lowering.packFunctionResults(resultTypes)))
         return this->matchFailure();
@@ -1076,11 +1140,26 @@ struct MemRefCastOpLowering : public LLVMLegalizationPattern<MemRefCastOp> {
 
   PatternMatchResult match(Operation *op) const override {
     auto memRefCastOp = cast<MemRefCastOp>(op);
-    MemRefType sourceType =
+    Type srcType = memRefCastOp.getOperand()->getType();
-        memRefCastOp.getOperand()->getType().cast<MemRefType>();
+    Type dstType = memRefCastOp.getType();
-    MemRefType targetType = memRefCastOp.getType();
+
-    return (isSupportedMemRefType(targetType) &&
+    if (srcType.isa<MemRefType>() && dstType.isa<MemRefType>()) {
-            isSupportedMemRefType(sourceType))
+      MemRefType sourceType =
+          memRefCastOp.getOperand()->getType().cast<MemRefType>();
+      MemRefType targetType = memRefCastOp.getType().cast<MemRefType>();
+      return (isSupportedMemRefType(targetType) &&
+              isSupportedMemRefType(sourceType))
+                 ? matchSuccess()
+                 : matchFailure();
+    }
+
+    // At least one of the operands is unranked type
+    assert(srcType.isa<UnrankedMemRefType>() ||
+           dstType.isa<UnrankedMemRefType>());
+
+    // Unranked to unranked cast is disallowed
+    return !(srcType.isa<UnrankedMemRefType>() &&
+             dstType.isa<UnrankedMemRefType>())
                ? matchSuccess()
                : matchFailure();
   }
@@ -1089,12 +1168,65 @@ struct MemRefCastOpLowering : public LLVMLegalizationPattern<MemRefCastOp> {
                ConversionPatternRewriter &rewriter) const override {
     auto memRefCastOp = cast<MemRefCastOp>(op);
     OperandAdaptor<MemRefCastOp> transformed(operands);
-    // memref_cast is defined for source and destination memref types with the
+
-    // same element type, same mappings, same address space and same rank.
+    auto srcType = memRefCastOp.getOperand()->getType();
-    // Therefore a simple bitcast suffices. If not it is undefined behavior.
+    auto dstType = memRefCastOp.getType();
     auto targetStructType = lowering.convertType(memRefCastOp.getType());
-    rewriter.replaceOpWithNewOp<LLVM::BitcastOp>(op, targetStructType,
+    auto loc = op->getLoc();
-                                                 transformed.source());
+
+    if (srcType.isa<MemRefType>() && dstType.isa<MemRefType>()) {
+      // memref_cast is defined for source and destination memref types with the
+      // same element type, same mappings, same address space and same rank.
+      // Therefore a simple bitcast suffices. If not it is undefined behavior.
+      rewriter.replaceOpWithNewOp<LLVM::BitcastOp>(op, targetStructType,
+                                                   transformed.source());
+    } else if (srcType.isa<MemRefType>() && dstType.isa<UnrankedMemRefType>()) {
+      // Casting ranked to unranked memref type
+      // Set the rank in the destination from the memref type
+      // Allocate space on the stack and copy the src memref decsriptor
+      // Set the ptr in the destination to the stack space
+      auto srcMemRefType = srcType.cast<MemRefType>();
+      int64_t rank = srcMemRefType.getRank();
+      // ptr = AllocaOp sizeof(MemRefDescriptor)
+      auto ptr = lowering.promoteOneMemRefDescriptor(loc, transformed.source(),
+                                                     rewriter);
+      // voidptr = BitCastOp srcType* to void*
+      auto voidPtr =
+          rewriter.create<LLVM::BitcastOp>(loc, getVoidPtrType(), ptr)
+              .getResult();
+      // rank = ConstantOp srcRank
+      auto rankVal = rewriter.create<LLVM::ConstantOp>(
+          loc, lowering.convertType(rewriter.getIntegerType(64)),
+          rewriter.getI64IntegerAttr(rank));
+      // undef = UndefOp
+      UnrankedMemRefDescriptor memRefDesc =
+          UnrankedMemRefDescriptor::undef(rewriter, loc, targetStructType);
+      // d1 = InsertValueOp undef, rank, 0
+      memRefDesc.setRank(rewriter, loc, rankVal);
+      // d2 = InsertValueOp d1, voidptr, 1
+      memRefDesc.setMemRefDescPtr(rewriter, loc, voidPtr);
+      rewriter.replaceOp(op, (Value *)memRefDesc);
+
+    } else if (srcType.isa<UnrankedMemRefType>() && dstType.isa<MemRefType>()) {
+      // Casting from unranked type to ranked.
+      // The operation is assumed to be doing a correct cast. If the destination
+      // type mismatches the unranked the type, it is undefined behavior.
+      UnrankedMemRefDescriptor memRefDesc(transformed.source());
+      // ptr = ExtractValueOp src, 1
+      auto ptr = memRefDesc.memRefDescPtr(rewriter, loc);
+      // castPtr = BitCastOp i8* to structTy*
+      auto castPtr =
+          rewriter
+              .create<LLVM::BitcastOp>(
+                  loc, targetStructType.cast<LLVM::LLVMType>().getPointerTo(),
+                  ptr)
+              .getResult();
+      // struct = LoadOp castPtr
+      auto loadOp = rewriter.create<LLVM::LoadOp>(loc, castPtr);
+      rewriter.replaceOp(op, loadOp.getResult());
+    } else {
+      llvm_unreachable("Unsuppored unranked memref to unranked memref cast");
+    }
   }
 };
 
@@ -1896,7 +2028,8 @@ SmallVector<Value *, 4> LLVMTypeConverter::promoteMemRefDescriptors(
   for (auto it : llvm::zip(opOperands, operands)) {
     auto *operand = std::get<0>(it);
     auto *llvmOperand = std::get<1>(it);
-    if (!operand->getType().isa<MemRefType>()) {
+    if (!operand->getType().isa<MemRefType>() &&
+        !operand->getType().isa<UnrankedMemRefType>()) {
       promotedOperands.push_back(operand);
       continue;
     }

mlir/lib/Dialect/StandardOps/Ops.cpp

@@ -1769,46 +1769,70 @@ bool MemRefCastOp::areCastCompatible(Type a, Type b) {
   auto aT = a.dyn_cast<MemRefType>();
   auto bT = b.dyn_cast<MemRefType>();
 
-  if (!aT || !bT)
+  auto uaT = a.dyn_cast<UnrankedMemRefType>();
-    return false;
+  auto ubT = b.dyn_cast<UnrankedMemRefType>();
-  if (aT.getElementType() != bT.getElementType())
-    return false;
-  if (aT.getAffineMaps() != bT.getAffineMaps()) {
-    int64_t aOffset, bOffset;
-    SmallVector<int64_t, 4> aStrides, bStrides;
-    if (failed(getStridesAndOffset(aT, aStrides, aOffset)) ||
-        failed(getStridesAndOffset(bT, bStrides, bOffset)) ||
-        aStrides.size() != bStrides.size())
-      return false;
 
-    // Strides along a dimension/offset are compatible if the value in the
+  if (aT && bT) {
-    // source memref is static and the value in the target memref is the
+    if (aT.getElementType() != bT.getElementType())
-    // same. They are also compatible if either one is dynamic (see description
-    // of MemRefCastOp for details).
-    auto checkCompatible = [](int64_t a, int64_t b) {
-      return (a == MemRefType::getDynamicStrideOrOffset() ||
-              b == MemRefType::getDynamicStrideOrOffset() || a == b);
-    };
-    if (!checkCompatible(aOffset, bOffset))
       return false;
-    for (auto aStride : enumerate(aStrides))
+    if (aT.getAffineMaps() != bT.getAffineMaps()) {
-      if (!checkCompatible(aStride.value(), bStrides[aStride.index()]))
+      int64_t aOffset, bOffset;
+      SmallVector<int64_t, 4> aStrides, bStrides;
+      if (failed(getStridesAndOffset(aT, aStrides, aOffset)) ||
+          failed(getStridesAndOffset(bT, bStrides, bOffset)) ||
+          aStrides.size() != bStrides.size())
         return false;
-  }
-  if (aT.getMemorySpace() != bT.getMemorySpace())
-    return false;
 
-  // They must have the same rank, and any specified dimensions must match.
+      // Strides along a dimension/offset are compatible if the value in the
-  if (aT.getRank() != bT.getRank())
+      // source memref is static and the value in the target memref is the
-    return false;
+      // same. They are also compatible if either one is dynamic (see
-
+      // description of MemRefCastOp for details).
-  for (unsigned i = 0, e = aT.getRank(); i != e; ++i) {
+      auto checkCompatible = [](int64_t a, int64_t b) {
-    int64_t aDim = aT.getDimSize(i), bDim = bT.getDimSize(i);
+        return (a == MemRefType::getDynamicStrideOrOffset() ||
-    if (aDim != -1 && bDim != -1 && aDim != bDim)
+                b == MemRefType::getDynamicStrideOrOffset() || a == b);
+      };
+      if (!checkCompatible(aOffset, bOffset))
+        return false;
+      for (auto aStride : enumerate(aStrides))
+        if (!checkCompatible(aStride.value(), bStrides[aStride.index()]))
+          return false;
+    }
+    if (aT.getMemorySpace() != bT.getMemorySpace())
       return false;
+
+    // They must have the same rank, and any specified dimensions must match.
+    if (aT.getRank() != bT.getRank())
+      return false;
+
+    for (unsigned i = 0, e = aT.getRank(); i != e; ++i) {
+      int64_t aDim = aT.getDimSize(i), bDim = bT.getDimSize(i);
+      if (aDim != -1 && bDim != -1 && aDim != bDim)
+        return false;
+    }
+    return true;
+  } else {
+    if (!aT && !uaT)
+      return false;
+    if (!bT && !ubT)
+      return false;
+    // Unranked to unranked casting is unsupported
+    if (uaT && ubT)
+      return false;
+
+    auto aEltType = (aT) ? aT.getElementType() : uaT.getElementType();
+    auto bEltType = (bT) ? bT.getElementType() : ubT.getElementType();
+    if (aEltType != bEltType)
+      return false;
+
+    auto aMemSpace = (aT) ? aT.getMemorySpace() : uaT.getMemorySpace();
+    auto bMemSpace = (bT) ? bT.getMemorySpace() : ubT.getMemorySpace();
+    if (aMemSpace != bMemSpace)
+      return false;
+
+    return true;
   }
 
-  return true;
+  return false;
 }
 
 OpFoldResult MemRefCastOp::fold(ArrayRef<Attribute> operands) {

mlir/lib/IR/AsmPrinter.cpp

@@ -1086,6 +1086,13 @@ void ModulePrinter::printType(Type type) {
     os << '>';
     return;
   }
+  case StandardTypes::UnrankedMemRef: {
+    auto v = type.cast<UnrankedMemRefType>();
+    os << "memref<*x";
+    printType(v.getElementType());
+    os << '>';
+    return;
+  }
   case StandardTypes::Complex:
     os << "complex<";
     printType(type.cast<ComplexType>().getElementType());

mlir/lib/IR/MLIRContext.cpp

@@ -90,8 +90,8 @@ struct BuiltinDialect : public Dialect {
                   UnknownLoc>();
 
     addTypes<ComplexType, FloatType, FunctionType, IndexType, IntegerType,
-             MemRefType, NoneType, OpaqueType, RankedTensorType, TupleType,
+             MemRefType, UnrankedMemRefType, NoneType, OpaqueType,
-             UnrankedTensorType, VectorType>();
+             RankedTensorType, TupleType, UnrankedTensorType, VectorType>();
 
     // TODO: These operations should be moved to a different dialect when they
     // have been fully decoupled from the core.

mlir/lib/IR/StandardTypes.cpp

@@ -390,6 +390,37 @@ ArrayRef<AffineMap> MemRefType::getAffineMaps() const {
 
 unsigned MemRefType::getMemorySpace() const { return getImpl()->memorySpace; }
 
+//===----------------------------------------------------------------------===//
+// UnrankedMemRefType
+//===----------------------------------------------------------------------===//
+
+UnrankedMemRefType UnrankedMemRefType::get(Type elementType,
+                                           unsigned memorySpace) {
+  return Base::get(elementType.getContext(), StandardTypes::UnrankedMemRef,
+                   elementType, memorySpace);
+}
+
+UnrankedMemRefType UnrankedMemRefType::getChecked(Type elementType,
+                                                  unsigned memorySpace,
+                                                  Location location) {
+  return Base::getChecked(location, elementType.getContext(),
+                          StandardTypes::UnrankedMemRef, elementType,
+                          memorySpace);
+}
+
+unsigned UnrankedMemRefType::getMemorySpace() const {
+  return getImpl()->memorySpace;
+}
+
+LogicalResult UnrankedMemRefType::verifyConstructionInvariants(
+    llvm::Optional<Location> loc, MLIRContext *context, Type elementType,
+    unsigned memorySpace) {
+  // Check that memref is formed from allowed types.
+  if (!elementType.isIntOrFloat() && !elementType.isa<VectorType>())
+    return emitOptionalError(*loc, "invalid memref element type");
+  return success();
+}
+
 /// Given MemRef `sizes` that are either static or dynamic, returns the
 /// canonical "contiguous" strides AffineExpr. Strides are multiplicative and
 /// once a dynamic dimension is encountered, all canonical strides become

mlir/lib/IR/TypeDetail.h

@@ -119,8 +119,8 @@ struct FunctionTypeStorage : public TypeStorage {
 
 /// Shaped Type Storage.
 struct ShapedTypeStorage : public TypeStorage {
-  ShapedTypeStorage(Type elementType, unsigned subclassData = 0)
+  ShapedTypeStorage(Type elementTy, unsigned subclassData = 0)
-      : TypeStorage(subclassData), elementType(elementType) {}
+      : TypeStorage(subclassData), elementType(elementTy) {}
 
   /// The hash key used for uniquing.
   using KeyTy = Type;
@@ -252,6 +252,31 @@ struct MemRefTypeStorage : public ShapedTypeStorage {
   const unsigned memorySpace;
 };
 
+/// Unranked MemRef is a MemRef with unknown rank.
+/// Only element type and memory space are known
+struct UnrankedMemRefTypeStorage : public ShapedTypeStorage {
+
+  UnrankedMemRefTypeStorage(Type elementTy, const unsigned memorySpace)
+      : ShapedTypeStorage(elementTy), memorySpace(memorySpace) {}
+
+  /// The hash key used for uniquing.
+  using KeyTy = std::tuple<Type, unsigned>;
+  bool operator==(const KeyTy &key) const {
+    return key == KeyTy(elementType, memorySpace);
+  }
+
+  /// Construction.
+  static UnrankedMemRefTypeStorage *construct(TypeStorageAllocator &allocator,
+                                              const KeyTy &key) {
+
+    // Initialize the memory using placement new.
+    return new (allocator.allocate<UnrankedMemRefTypeStorage>())
+        UnrankedMemRefTypeStorage(std::get<0>(key), std::get<1>(key));
+  }
+  /// Memory space in which data referenced by memref resides.
+  const unsigned memorySpace;
+};
+
 /// Complex Type Storage.
 struct ComplexTypeStorage : public TypeStorage {
   ComplexTypeStorage(Type elementType) : elementType(elementType) {}

mlir/lib/Parser/Parser.cpp

@@ -1054,8 +1054,13 @@ ParseResult Parser::parseStridedLayout(int64_t &offset,
 
 /// Parse a memref type.
 ///
-///   memref-type ::= `memref` `<` dimension-list-ranked type
+///   memref-type ::= ranked-memref-type | unranked-memref-type
-///                   (`,` semi-affine-map-composition)? (`,` memory-space)? `>`
+///
+///   ranked-memref-type ::= `memref` `<` dimension-list-ranked type
+///                          (`,` semi-affine-map-composition)? (`,`
+///                          memory-space)? `>`
+///
+///   unranked-memref-type ::= `memref` `<*x` type (`,` memory-space)? `>`
 ///
 ///   semi-affine-map-composition ::= (semi-affine-map `,` )* semi-affine-map
 ///   memory-space ::= integer-literal /* | TODO: address-space-id */
@@ -1066,9 +1071,20 @@ Type Parser::parseMemRefType() {
   if (parseToken(Token::less, "expected '<' in memref type"))
     return nullptr;
 
+  bool isUnranked;
   SmallVector<int64_t, 4> dimensions;
-  if (parseDimensionListRanked(dimensions))
+
-    return nullptr;
+  if (consumeIf(Token::star)) {
+    // This is an unranked memref type.
+    isUnranked = true;
+    if (parseXInDimensionList())
+      return nullptr;
+
+  } else {
+    isUnranked = false;
+    if (parseDimensionListRanked(dimensions))
+      return nullptr;
+  }
 
   // Parse the element type.
   auto typeLoc = getToken().getLoc();
@@ -1093,6 +1109,8 @@ Type Parser::parseMemRefType() {
       consumeToken(Token::integer);
       parsedMemorySpace = true;
     } else {
+      if (isUnranked)
+        return emitError("cannot have affine map for unranked memref type");
       if (parsedMemorySpace)
         return emitError("expected memory space to be last in memref type");
       if (getToken().is(Token::kw_offset)) {
@@ -1131,6 +1149,10 @@ Type Parser::parseMemRefType() {
       return nullptr;
   }
 
+  if (isUnranked)
+    return UnrankedMemRefType::getChecked(elementType, memorySpace,
+                                          getEncodedSourceLocation(typeLoc));
+
   return MemRefType::getChecked(dimensions, elementType, affineMapComposition,
                                 memorySpace, getEncodedSourceLocation(typeLoc));
 }

mlir/test/Conversion/StandardToLLVM/convert-memref-ops.mlir

@@ -371,6 +371,30 @@ func @memref_cast_mixed_to_mixed(%mixed : memref<42x?xf32>) {
   return
 }
 
+// CHECK-LABEL: func @memref_cast_ranked_to_unranked
+func @memref_cast_ranked_to_unranked(%arg : memref<42x2x?xf32>) {
+// CHECK-NEXT: %[[ld:.*]] = llvm.load %{{.*}} : !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }*">
+// CHECK-DAG:  %[[c:.*]] = llvm.mlir.constant(1 : index) : !llvm.i64
+// CHECK-DAG:  %[[p:.*]] = llvm.alloca %[[c]] x !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }"> : (!llvm.i64) -> !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }*">
+// CHECK-DAG:  llvm.store %[[ld]], %[[p]] : !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }*">
+// CHECK-DAG:  %[[p2:.*]] = llvm.bitcast %2 : !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }*"> to !llvm<"i8*">
+// CHECK-DAG:  %[[r:.*]] = llvm.mlir.constant(3 : i64) : !llvm.i64
+// CHECK    :  llvm.mlir.undef : !llvm<"{ i64, i8* }">
+// CHECK-DAG:  llvm.insertvalue %[[r]], %{{.*}}[0] : !llvm<"{ i64, i8* }">
+// CHECK-DAG:  llvm.insertvalue %[[p2]], %{{.*}}[1] : !llvm<"{ i64, i8* }">
+  %0 = memref_cast %arg : memref<42x2x?xf32> to memref<*xf32>
+  return
+}
+
+// CHECK-LABEL: func @memref_cast_unranked_to_ranked
+func @memref_cast_unranked_to_ranked(%arg : memref<*xf32>) {
+// CHECK: %[[ld:.*]] = llvm.load %{{.*}} : !llvm<"{ i64, i8* }*">
+// CHECK-NEXT: %[[p:.*]] = llvm.extractvalue %[[ld]][1] : !llvm<"{ i64, i8* }">
+// CHECK-NEXT: llvm.bitcast %[[p]] : !llvm<"i8*"> to !llvm<"{ float*, float*, i64, [4 x i64], [4 x i64] }*">
+  %0 = memref_cast %arg : memref<*xf32> to memref<?x?x10x2xf32>
+  return
+}
+
 // CHECK-LABEL: func @mixed_memref_dim(%arg0: !llvm<"{ float*, float*, i64, [5 x i64], [5 x i64] }*">) {
 func @mixed_memref_dim(%mixed : memref<42x?x?x13x?xf32>) {
 //  CHECK-NEXT:  %[[ld:.*]] = llvm.load %{{.*}} : !llvm<"{ float*, float*, i64, [5 x i64], [5 x i64] }*">

mlir/test/IR/core-ops.mlir

@@ -565,6 +565,11 @@ func @memref_cast(%arg0: memref<4xf32>, %arg1 : memref<?xf32>, %arg2 : memref<64
   // CHECK: {{%.*}} = memref_cast {{%.*}} : memref<64x16x4xf32, #[[BASE_MAP3]]> to memref<64x16x4xf32, #[[BASE_MAP0]]>
   %3 = memref_cast %2 : memref<64x16x4xf32, offset: ?, strides: [?, ?, ?]> to memref<64x16x4xf32, offset: 0, strides: [64, 4, 1]>
 
+  // CHECK: memref_cast %{{.*}} : memref<4xf32> to memref<*xf32>
+  %4 = memref_cast %1 : memref<4xf32> to memref<*xf32>
+
+  // CHECK: memref_cast %{{.*}} : memref<*xf32> to memref<4xf32>
+  %5 = memref_cast %4 : memref<*xf32> to memref<4xf32>
   return
 }
 

mlir/test/IR/invalid-ops.mlir

@@ -978,3 +978,34 @@ func @invalid_memref_cast(%arg0 : memref<12x4x16xf32, offset:0, strides:[64, 16,
   %0 = memref_cast %arg0 : memref<12x4x16xf32, offset:0, strides:[64, 16, 1]> to memref<12x4x16xf32, offset:16, strides:[64, 16, 1]>
   return
 }
+
+// -----
+
+// incompatible element types
+func @invalid_memref_cast() {
+  %0 = alloc() : memref<2x5xf32, 0>
+  // expected-error@+1 {{operand type 'memref<2x5xf32>' and result type 'memref<*xi32>' are cast incompatible}}
+  %1 = memref_cast %0 : memref<2x5xf32, 0> to memref<*xi32> 
+  return
+}
+
+// -----
+
+// incompatible memory space
+func @invalid_memref_cast() {
+  %0 = alloc() : memref<2x5xf32, 0>
+  // expected-error@+1 {{operand type 'memref<2x5xf32>' and result type 'memref<*xf32>' are cast incompatible}}
+  %1 = memref_cast %0 : memref<2x5xf32, 0> to memref<*xf32, 1> 
+  return
+}
+
+// -----
+
+// unranked to unranked
+func @invalid_memref_cast() {
+  %0 = alloc() : memref<2x5xf32, 0>
+  %1 = memref_cast %0 : memref<2x5xf32, 0> to memref<*xf32, 0> 
+  // expected-error@+1 {{operand type 'memref<*xf32>' and result type 'memref<*xf32>' are cast incompatible}}
+  %2 = memref_cast %1 : memref<*xf32, 0> to memref<*xf32, 0> 
+  return
+}

mlir/test/mlir-cpu-runner/include/mlir_runner_utils.h

@@ -78,6 +78,12 @@ template <typename T> struct StridedMemRefType<T, 0> {
   int64_t offset;
 };
 
+// Unranked MemRef
+struct UnrankedMemRefType {
+  int64_t rank;
+  void *descriptor;
+};
+
 template <typename StreamType, typename T, int N>
 void printMemRefMetaData(StreamType &os, StridedMemRefType<T, N> &V) {
   static_assert(N > 0, "Expected N > 0");
@@ -97,6 +103,15 @@ void printMemRefMetaData(StreamType &os, StridedMemRefType<T, 0> &V) {
      << " offset = " << V.offset;
 }
 
+extern "C" MLIR_RUNNER_UTILS_EXPORT void
+print_memref_f32(UnrankedMemRefType *M);
+
+template <typename StreamType>
+void printUnrankedMemRefMetaData(StreamType &os, UnrankedMemRefType &V) {
+  os << "Unranked Memref rank = " << V.rank << " "
+     << "descriptor@ = " << reinterpret_cast<float *>(V.descriptor) << " ";
+}
+
 extern "C" MLIR_RUNNER_UTILS_EXPORT void
 print_memref_0d_f32(StridedMemRefType<float, 0> *M);
 extern "C" MLIR_RUNNER_UTILS_EXPORT void

mlir/test/mlir-cpu-runner/mlir_runner_utils.cpp

@@ -148,15 +148,41 @@ template <typename T, int N> void printMemRef(StridedMemRefType<T, N> &M) {
   std::cout << std::endl;
 }
 
-template <typename T> void printZeroDMemRef(StridedMemRefType<T, 0> &M) {
+template <typename T> void printMemRef(StridedMemRefType<T, 0> &M) {
   std::cout << "\nMemref base@ = " << M.data << " rank = " << 0
             << " offset = " << M.offset << " data = [";
   MemRefDataPrinter<T, 0>::print(std::cout, M.data, 0, M.offset);
   std::cout << "]" << std::endl;
 }
 
+extern "C" void
+print_memref_vector_4x4xf32(StridedMemRefType<Vector2D<4, 4, float>, 2> *M) {
+  printMemRef(*M);
+}
+
+extern "C" void print_memref_f32(UnrankedMemRefType *M) {
+  printUnrankedMemRefMetaData(std::cout, *M);
+  int rank = M->rank;
+  void *ptr = M->descriptor;
+
+#define MEMREF_CASE(RANK)                                                      \
+  case RANK:                                                                   \
+    printMemRef(*(static_cast<StridedMemRefType<float, RANK> *>(ptr)));        \
+    break
+
+  switch (rank) {
+    MEMREF_CASE(0);
+    MEMREF_CASE(1);
+    MEMREF_CASE(2);
+    MEMREF_CASE(3);
+    MEMREF_CASE(4);
+  default:
+    assert(0 && "Unsupported rank to print");
+  }
+}
+
 extern "C" void print_memref_0d_f32(StridedMemRefType<float, 0> *M) {
-  printZeroDMemRef(*M);
+  printMemRef(*M);
 }
 extern "C" void print_memref_1d_f32(StridedMemRefType<float, 1> *M) {
   printMemRef(*M);
@@ -170,8 +196,3 @@ extern "C" void print_memref_3d_f32(StridedMemRefType<float, 3> *M) {
 extern "C" void print_memref_4d_f32(StridedMemRefType<float, 4> *M) {
   printMemRef(*M);
 }
-
-extern "C" void
-print_memref_vector_4x4xf32(StridedMemRefType<Vector2D<4, 4, float>, 2> *M) {
-  printMemRef(*M);
-}

mlir/test/mlir-cpu-runner/unranked_memref.mlir

@@ -0,0 +1,43 @@
+// RUN: mlir-opt %s -convert-linalg-to-loops -convert-linalg-to-llvm -convert-std-to-llvm | mlir-cpu-runner -e main -entry-point-result=void -shared-libs=%linalg_test_lib_dir/libmlir_runner_utils%shlibext,%linalg_test_lib_dir/libcblas%shlibext,%linalg_test_lib_dir/libcblas_interface%shlibext | FileCheck %s
+
+// CHECK: rank = 2
+// CHECK-SAME: sizes = [10, 3]
+// CHECK-SAME: strides = [3, 1]
+// CHECK-COUNT-10: [10, 10, 10]
+//
+// CHECK: rank = 2
+// CHECK-SAME: sizes = [10, 3]
+// CHECK-SAME: strides = [3, 1]
+// CHECK-COUNT-10: [5, 5, 5]
+//
+// CHECK: rank = 2
+// CHECK-SAME: sizes = [10, 3]
+// CHECK-SAME: strides = [3, 1]
+// CHECK-COUNT-10: [2, 2, 2]
+func @main() -> () {
+    %A = alloc() : memref<10x3xf32, 0>
+    %f2 = constant 2.00000e+00 : f32
+    %f5 = constant 5.00000e+00 : f32
+    %f10 = constant 10.00000e+00 : f32
+
+    %V = memref_cast %A : memref<10x3xf32, 0> to memref<?x?xf32>
+    linalg.fill(%V, %f10) : memref<?x?xf32, 0>, f32
+    %U = memref_cast %A : memref<10x3xf32, 0> to memref<*xf32>
+    call @print_memref_f32(%U) : (memref<*xf32>) -> ()
+
+    %V2 = memref_cast %U : memref<*xf32> to memref<?x?xf32>
+    linalg.fill(%V2, %f5) : memref<?x?xf32, 0>, f32
+    %U2 = memref_cast %V2 : memref<?x?xf32, 0> to memref<*xf32>
+    call @print_memref_f32(%U2) : (memref<*xf32>) -> ()
+
+    %V3 = memref_cast %V2 : memref<?x?xf32> to memref<*xf32>
+    %V4 = memref_cast %V3 : memref<*xf32> to memref<?x?xf32>
+    linalg.fill(%V4, %f2) : memref<?x?xf32, 0>, f32
+    %U3 = memref_cast %V2 : memref<?x?xf32> to memref<*xf32>
+    call @print_memref_f32(%U3) : (memref<*xf32>) -> ()
+
+    dealloc %A : memref<10x3xf32, 0>
+    return
+}
+
+func @print_memref_f32(memref<*xf32>)

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

@@ -7,7 +7,8 @@ func @print_0d() {
   %f = constant 2.00000e+00 : f32
   %A = alloc() : memref<f32>
   store %f, %A[]: memref<f32>
-  call @print_memref_0d_f32(%A): (memref<f32>) -> ()
+  %U = memref_cast %A :  memref<f32> to memref<*xf32>
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
   dealloc %A : memref<f32>
   return
 }
@@ -18,7 +19,8 @@ func @print_1d() {
   %A = alloc() : memref<16xf32>
   %B = memref_cast %A: memref<16xf32> to memref<?xf32>
   linalg.fill(%B, %f) : memref<?xf32>, f32
-  call @print_memref_1d_f32(%B): (memref<?xf32>) -> ()
+  %U = memref_cast %B :  memref<?xf32> to memref<*xf32>
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
   dealloc %A : memref<16xf32>
   return
 }
@@ -34,8 +36,8 @@ func @print_3d() {
 
   %c2 = constant 2 : index
   store %f4, %B[%c2, %c2, %c2]: memref<?x?x?xf32>
-
+  %U = memref_cast %B : memref<?x?x?xf32> to memref<*xf32>
-  call @print_memref_3d_f32(%B): (memref<?x?x?xf32>) -> ()
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
   dealloc %A : memref<3x4x5xf32>
   return
 }
@@ -46,10 +48,7 @@ func @print_3d() {
 //    PRINT-3D-NEXT: 2,    2,    4,    2,    2
 //    PRINT-3D-NEXT: 2,    2,    2,    2,    2
 
-func @print_memref_0d_f32(memref<f32>)
+func @print_memref_f32(memref<*xf32>)
-func @print_memref_1d_f32(memref<?xf32>)
-func @print_memref_3d_f32(memref<?x?x?xf32>)
-
 
 !vector_type_C = type vector<4x4xf32>
 !matrix_type_CC = type memref<1x1x!vector_type_C>

mlir/test/mlir-cuda-runner/all-reduce-op.mlir

@@ -22,9 +22,10 @@ func @main() {
     store %sum, %kernel_dst[%tz, %ty, %tx] : memref<?x?x?xf32>
     gpu.return
   }
-  call @print_memref_3d_f32(%dst) : (memref<?x?x?xf32>) -> ()
+  %U = memref_cast %dst : memref<?x?x?xf32> to memref<*xf32>
+  call @print_memref_f32(%U) : (memref<*xf32>) -> ()
   return
 }
 
 func @mcuMemHostRegisterMemRef3dFloat(%ptr : memref<?x?x?xf32>)
-func @print_memref_3d_f32(%ptr : memref<?x?x?xf32>)
+func @print_memref_f32(%ptr : memref<*xf32>)

mlir/test/mlir-cuda-runner/all-reduce-region.mlir

@@ -20,9 +20,10 @@ func @main() {
     store %res, %kernel_dst[%tx] : memref<?xf32>
     gpu.return
   }
-  call @print_memref_1d_f32(%dst) : (memref<?xf32>) -> ()
+  %U = memref_cast %dst : memref<?xf32> to memref<*xf32>
+  call @print_memref_f32(%U) : (memref<*xf32>) -> ()
   return
 }
 
 func @mcuMemHostRegisterMemRef1dFloat(%ptr : memref<?xf32>)
-func @print_memref_1d_f32(memref<?xf32>)
+func @print_memref_f32(memref<*xf32>)