[Mlir-commits] [mlir] [mlir][tosa] Combine unranked/ranked tensor types into single type (PR #209737)

[Sayan Saha]

================
@@ -231,68 +223,48 @@ class Tosa_TensorOfOrNone<list<Type> allowedTypes, string description = ""> :
 // Tensor types with constrained ranks.
 //===----------------------------------------------------------------------===//
 
-def Tosa_Rank0Tensor : TosaTensorRankOf<[Tosa_AnyNumber], [0]>;
-
-def Tosa_ScalarTensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int, AnyFloat]>, TosaScalarTensorOf<[Tosa_AnyNumber], [1]>]>;
-def Tosa_ScalarInt8Tensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int8]>, TosaScalarTensorOf<[Tosa_Int8], [1]>]>;
-def Tosa_ScalarIntOrFloatTensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int, AnyFloat]>, TosaScalarTensorOf<[Tosa_Int, AnyFloat], [1]>]>;
-def Tosa_ScalarInt32Tensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int32]>, TosaScalarTensorOf<[Tosa_Int32], [1]>]>;
-
-// We include unranked tensors as a supported type for all possible tosa
-// Tensors as unranked does not guarantee invalid. If unranked tensors exist
-// they should be shape propagate used Tosa's shape inference pass and verified
-// to not include any remaining unranked tensors.
-def Tosa_UnrankedTensor : TosaUnrankedTensorOf<[Tosa_AnyNumber]>;
+def Tosa_ScalarTensor : TosaScalarTensorOf<[Tosa_AnyNumber], [1]>;
+def Tosa_ScalarInt8Tensor : TosaScalarTensorOf<[Tosa_Int8], [1]>;
+def Tosa_ScalarIntOrFloatTensor : TosaScalarTensorOf<[Tosa_Int, AnyFloat], [1]>;
+def Tosa_ScalarInt32Tensor : TosaScalarTensorOf<[Tosa_Int32], [1]>;
 
-def Tosa_Tensor1D : AnyTypeOf<[Tosa_UnrankedTensor, TosaTensorRankOf<[Tosa_AnyNumber], [1]>], "1-d tosa-conformant tensor", "::mlir::TensorType">;
-def Tosa_Tensor2D : AnyTypeOf<[Tosa_UnrankedTensor, TosaTensorRankOf<[Tosa_AnyNumber], [2]>], "2-d tosa-conformant tensor", "::mlir::TensorType">;
-def Tosa_Tensor3D : AnyTypeOf<[Tosa_UnrankedTensor, TosaTensorRankOf<[Tosa_AnyNumber], [3]>], "3-d tosa-conformant tensor", "::mlir::TensorType">;
-def Tosa_Tensor4D : AnyTypeOf<[Tosa_UnrankedTensor, TosaTensorRankOf<[Tosa_AnyNumber], [4]>], "4-d tosa-conformant tensor", "::mlir::TensorType">;
-def Tosa_Tensor5D : AnyTypeOf<[Tosa_UnrankedTensor, TosaTensorRankOf<[Tosa_AnyNumber], [5]>], "5-d tosa-conformant tensor", "::mlir::TensorType">;
+def Tosa_Tensor1D : TosaTensorRankOf<[Tosa_AnyNumber], [1]>;
+def Tosa_Tensor2D : TosaTensorRankOf<[Tosa_AnyNumber], [2]>;
+def Tosa_Tensor3D : TosaTensorRankOf<[Tosa_AnyNumber], [3]>;
+def Tosa_Tensor4D : TosaTensorRankOf<[Tosa_AnyNumber], [4]>;
+def Tosa_Tensor5D : TosaTensorRankOf<[Tosa_AnyNumber], [5]>;
 
 // 1D tensor of specific types
-def Tosa_1DInt8Tensor : 1DTensorOf<[Tosa_Int8]>;
-def Tosa_1DInt16Or32Tensor : 1DTensorOf<[Tosa_Int16Or32]>;
+def Tosa_1DInt8Tensor : TosaTensorRankOf<[Tosa_Int8], [1]>;
----------------
sahas3 wrote:

This may cause a breaking change in downstream consumers -- I think before the type will be `RankedTensorType` but now it will be `TensorType`? I think we have some code that implicitly casts to `TypedValue<RankedTensorType>` which won't work anymore. Easily fixable but I don't know if there's some announcement to be made about this change.

https://github.com/llvm/llvm-project/pull/209737