[MLIR][Linalg] Add more arith named ops to linalg

[rengolin]

Summary

Following up the 'add' named op, here are the remaining basic arithmetic
and maths, including a 'div_unsigned' for integer unsigned values. In the
same pattern as 'matmul_unsigned', the simply named 'div' assumes signed
values and the '_unsigned' variation handles the unsigned values.

It's a bit odd, but there doesn't seem to be a easy way to restrict to
specific types to make 'div_unsigned' only work with integers in the
structured ops framework.

Same as 'add', these have strict semantics regarding casts.

Unary math ops will need some massaging, so I split these ones for now
as I continue working on them.

[mehdi_amini]

Seems like this is broken in CI? https://lab.llvm.org/buildbot/#/builders/61/builds/45549

[rengolin]

Yes, I already contacted the bot owner. Happy to do some investigation on my side, but I'm not sure how.

[NathanielMcVicar]

It also appears to have broken Windows and all the other bots. https://lab.llvm.org/buildbot/#/builders/13/builds/37348/steps/7/logs/FAIL__MLIR__emit_matmul_py

[rengolin]

Oh well, that's a clear revert.

[rengolin]

Ok, reverted. I'll try to reproduce the error and try again tomorrow.

[rengolin]

Found it, I was missing the following arguments to my CMAKE:

-DMLIR_INCLUDE_INTEGRATION_TESTS=ON -DMLIR_ENABLE_BINDINGS_PYTHON=ON

Am looking at the errors and will try again once green.

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

 ]> {
   let genSpecializedAttr = 0;
   let cppNamespace = "::mlir::linalg";
 }
 def BinaryFn : I32EnumAttr<"BinaryFn", "", [
   I32EnumAttrCase<"add", 0>,
   I32EnumAttrCase<"sub", 1>,
   I32EnumAttrCase<"mul", 2>,
-  I32EnumAttrCase<"max_signed", 3>,
-  I32EnumAttrCase<"min_signed", 4>,
-  I32EnumAttrCase<"max_unsigned", 5>,
-  I32EnumAttrCase<"min_unsigned", 6>
+  I32EnumAttrCase<"div", 3>,
+  I32EnumAttrCase<"div_unsigned", 4>,
+  I32EnumAttrCase<"max_signed", 5>,
+  I32EnumAttrCase<"min_signed", 6>,
+  I32EnumAttrCase<"max_unsigned", 7>,
+  I32EnumAttrCase<"min_unsigned", 8>
 ]> {
   let genSpecializedAttr = 0;
   let cppNamespace = "::mlir::linalg";
 }
 def TypeFn : I32EnumAttr<"TypeFn", "", [
   I32EnumAttrCase<"cast_signed", 0>,
   I32EnumAttrCase<"cast_unsigned", 1>
 ]> {
   let genSpecializedAttr = 0;
   let cppNamespace = "::mlir::linalg";
 }
 
 #endif // LINALG_ENUMS

mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml

value: !ScalarExpression
         fn_name: add
         operands:
         - !ScalarExpression
           scalar_arg: lhs
         - !ScalarExpression
           scalar_arg: rhs
 --- !LinalgOpConfig
 metadata: !LinalgOpMetadata
+  name: sub
+  cpp_class_name: SubOp
+  doc: |-
+    Subtracts two tensors elementwise.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.sub` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: lhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: rhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: out
+    kind: output_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+  iterator_types: []
+  assignments:
+  - !ScalarAssign
+    arg: out
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: sub
+        operands:
+        - !ScalarExpression
+          scalar_arg: lhs
+        - !ScalarExpression
+          scalar_arg: rhs
+--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+  name: mul
+  cpp_class_name: MulOp
+  doc: |-
+    Multiply two tensors elementwise.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.mul` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: lhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: rhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: out
+    kind: output_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+  iterator_types: []
+  assignments:
+  - !ScalarAssign
+    arg: out
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: mul
+        operands:
+        - !ScalarExpression
+          scalar_arg: lhs
+        - !ScalarExpression
+          scalar_arg: rhs
+--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+  name: div
+  cpp_class_name: DivOp
+  doc: |-
+    Divides the first tensor by the second tensor, elementwise. For integer
+    types, performs a signed division.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.div` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: lhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: rhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: out
+    kind: output_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+  iterator_types: []
+  assignments:
+  - !ScalarAssign
+    arg: out
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: div
+        operands:
+        - !ScalarExpression
+          scalar_arg: lhs
+        - !ScalarExpression
+          scalar_arg: rhs
+--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+  name: div_unsigned
+  cpp_class_name: DivUOp
+  doc: |-
+    Divides the first tensor by the second tensor, elementwise. For integer
+    types, performs an unsigned division.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.div` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: lhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: rhs
+    kind: input_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  - !LinalgOperandDefConfig
+    name: out
+    kind: output_tensor
+    type_var: T
+    shape_map: affine_map<() -> ()>
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+    - affine_map<() -> ()>
+  iterator_types: []
+  assignments:
+  - !ScalarAssign
+    arg: out
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: div_unsigned
+        operands:
+        - !ScalarExpression
+          scalar_arg: lhs
+        - !ScalarExpression
+          scalar_arg: rhs
+--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
   name: matmul
   cpp_class_name: MatmulOp
   doc: |-
     Performs a matrix multiplication of two 2D inputs.
 
     Numeric casting is performed on the operands to the inner multiply, promoting
     them to the same data type as the accumulator/output.
   implements:

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

Value buildBinaryFn(BinaryFn binaryFn, Value arg0, Value arg1) {
     case BinaryFn::mul:
       if (allComplex)
         return builder.create<complex::MulOp>(arg0.getLoc(), arg0, arg1);
       if (allFloatingPoint)
         return builder.create<arith::MulFOp>(arg0.getLoc(), arg0, arg1);
       if (allBool)
         return builder.create<arith::AndIOp>(arg0.getLoc(), arg0, arg1);
       return builder.create<arith::MulIOp>(arg0.getLoc(), arg0, arg1);
+    case BinaryFn::div:
+      if (allComplex)
+        return builder.create<complex::DivOp>(arg0.getLoc(), arg0, arg1);
+      if (allFloatingPoint)
+        return builder.create<arith::DivFOp>(arg0.getLoc(), arg0, arg1);
+      if (allBool)
+        llvm_unreachable("unsupported operation: div with bools");
+      return builder.create<arith::DivSIOp>(arg0.getLoc(), arg0, arg1);
+    case BinaryFn::div_unsigned:
+      if (allComplex)
+        return builder.create<complex::DivOp>(arg0.getLoc(), arg0, arg1);
+      if (allFloatingPoint)
+        return builder.create<arith::DivFOp>(arg0.getLoc(), arg0, arg1);
+      if (allBool)
+        llvm_unreachable("unsupported operation: div with bools");
+      return builder.create<arith::DivUIOp>(arg0.getLoc(), arg0, arg1);
     case BinaryFn::max_signed:
       assert(!allComplex);
       if (allFloatingPoint)
         return builder.create<arith::MaxFOp>(arg0.getLoc(), arg0, arg1);
       return builder.create<arith::MaxSIOp>(arg0.getLoc(), arg0, arg1);
     case BinaryFn::min_signed:
       assert(!allComplex);
       if (allFloatingPoint)

mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py

 
 
 @linalg_structured_op
 def add(
     lhs=TensorDef(T1),
     rhs=TensorDef(T1),
     O=TensorDef(T1, output=True),
 ):
-    """ Adds two tensors elementwise.
+    """Adds two tensors elementwise.
 
     The shapes and element types must be identical. The appropriate casts,
     broadcasts and reductions should be done previously to calling this op.
 
     This means reduction/broadcast/element cast semantics is explicit. Further
     passes can take that into account when lowering this code. For example,
     a `linalg.broadcast` + `linalg.add` sequence can be lowered to a
     `linalg.generic` with different affine maps for the two operands.
     """
     O[None] = lhs[None] + rhs[None]
 
 
 @linalg_structured_op
+def sub(
+    lhs=TensorDef(T1),
+    rhs=TensorDef(T1),
+    O=TensorDef(T1, output=True),
+):
+    """Subtracts two tensors elementwise.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.sub` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+    """
+    O[None] = lhs[None] - rhs[None]
+
+
+@linalg_structured_op
+def mul(
+    lhs=TensorDef(T1),
+    rhs=TensorDef(T1),
+    O=TensorDef(T1, output=True),
+):
+    """Multiplies two tensors elementwise.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.mul` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+    """
+    O[None] = lhs[None] * rhs[None]
+
+
+@linalg_structured_op
+def div(
+    lhs=TensorDef(T1),
+    rhs=TensorDef(T1),
+    O=TensorDef(T1, output=True),
+):
+    """Divides the first tensor by the second tensor, elementwise.
+
+    The shapes and element types must be identical. The appropriate casts,
+    broadcasts and reductions should be done previously to calling this op.
+
+    This means reduction/broadcast/element cast semantics is explicit. Further
+    passes can take that into account when lowering this code. For example,
+    a `linalg.broadcast` + `linalg.div` sequence can be lowered to a
+    `linalg.generic` with different affine maps for the two operands.
+    """
+    O[None] = lhs[None] / rhs[None]
+
+
+@linalg_structured_op
 def matmul(
     A=TensorDef(T1, S.M, S.K),
     B=TensorDef(T2, S.K, S.N),
     C=TensorDef(U, S.M, S.N, output=True),
     cast=TypeFnAttrDef(default=TypeFn.cast_signed),
 ):
     """Performs a matrix multiplication of two 2D inputs.
 

mlir/test/Dialect/Linalg/generalize-named-ops.mlir

 // CHECK-SAME: indexing_maps = [#[[MAP]], #[[MAP]], #[[MAP]]]
 // CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel"]}
 // CHECK-SAME:  ins(%[[LHS]], %[[RHS]] : memref<7x14x21xf32>, memref<7x14x21xf32>)
 // CHECK-SAME: outs(%[[OUT]] : memref<7x14x21xf32>)
 
 // CHECK:         ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
 // CHECK-NEXT:      %[[SUM:.+]] = arith.addf %[[BBARG0]], %[[BBARG1]] : f32
 // CHECK-NEXT:      linalg.yield %[[SUM]] : f32
+
+// -----
+
+func.func @generalize_sub(%lhs: memref<7x14x21xf32>, %rhs: memref<7x14x21xf32>,
+                          %out: memref<7x14x21xf32>) {
+  linalg.sub ins(%lhs, %rhs : memref<7x14x21xf32>, memref<7x14x21xf32>)
+             outs(%out : memref<7x14x21xf32>)
+  return
+}
+
+// CHECK-DAG: #[[MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
+
+// CHECK: func @generalize_sub
+// CHECK-SAME: (%[[LHS:.+]]: memref<7x14x21xf32>, %[[RHS:.+]]: memref<7x14x21xf32>,
+// CHECK-SAME:  %[[OUT:.+]]: memref<7x14x21xf32>)
+
+// CHECK: linalg.generic
+// CHECK-SAME: indexing_maps = [#[[MAP]], #[[MAP]], #[[MAP]]]
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel"]}
+// CHECK-SAME:  ins(%[[LHS]], %[[RHS]] : memref<7x14x21xf32>, memref<7x14x21xf32>)
+// CHECK-SAME: outs(%[[OUT]] : memref<7x14x21xf32>)
+
+// CHECK:         ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
+// CHECK-NEXT:      %[[SUM:.+]] = arith.subf %[[BBARG0]], %[[BBARG1]] : f32
+// CHECK-NEXT:      linalg.yield %[[SUM]] : f32
+
+// -----
+
+func.func @generalize_mul(%lhs: memref<7x14x21xf32>, %rhs: memref<7x14x21xf32>,
+                          %out: memref<7x14x21xf32>) {
+  linalg.mul ins(%lhs, %rhs : memref<7x14x21xf32>, memref<7x14x21xf32>)
+             outs(%out : memref<7x14x21xf32>)
+  return
+}
+
+// CHECK-DAG: #[[MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
+
+// CHECK: func @generalize_mul
+// CHECK-SAME: (%[[LHS:.+]]: memref<7x14x21xf32>, %[[RHS:.+]]: memref<7x14x21xf32>,
+// CHECK-SAME:  %[[OUT:.+]]: memref<7x14x21xf32>)
+
+// CHECK: linalg.generic
+// CHECK-SAME: indexing_maps = [#[[MAP]], #[[MAP]], #[[MAP]]]
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel"]}
+// CHECK-SAME:  ins(%[[LHS]], %[[RHS]] : memref<7x14x21xf32>, memref<7x14x21xf32>)
+// CHECK-SAME: outs(%[[OUT]] : memref<7x14x21xf32>)
+
+// CHECK:         ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
+// CHECK-NEXT:      %[[SUM:.+]] = arith.mulf %[[BBARG0]], %[[BBARG1]] : f32
+// CHECK-NEXT:      linalg.yield %[[SUM]] : f32
+
+// -----
+
+func.func @generalize_div(%lhs: memref<7x14x21xf32>, %rhs: memref<7x14x21xf32>,
+                          %out: memref<7x14x21xf32>) {
+  linalg.div ins(%lhs, %rhs : memref<7x14x21xf32>, memref<7x14x21xf32>)
+             outs(%out : memref<7x14x21xf32>)
+  return
+}
+
+// CHECK-DAG: #[[MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
+
+// CHECK: func @generalize_div
+// CHECK-SAME: (%[[LHS:.+]]: memref<7x14x21xf32>, %[[RHS:.+]]: memref<7x14x21xf32>,
+// CHECK-SAME:  %[[OUT:.+]]: memref<7x14x21xf32>)
+
+// CHECK: linalg.generic
+// CHECK-SAME: indexing_maps = [#[[MAP]], #[[MAP]], #[[MAP]]]
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel"]}
+// CHECK-SAME:  ins(%[[LHS]], %[[RHS]] : memref<7x14x21xf32>, memref<7x14x21xf32>)
+// CHECK-SAME: outs(%[[OUT]] : memref<7x14x21xf32>)
+
+// CHECK:         ^{{.+}}(%[[BBARG0:.+]]: f32, %[[BBARG1:.+]]: f32, %[[BBARG2:.+]]: f32)
+// CHECK-NEXT:      %[[SUM:.+]] = arith.divf %[[BBARG0]], %[[BBARG1]] : f32
+// CHECK-NEXT:      linalg.yield %[[SUM]] : f32
+
+// -----
+
+func.func @generalize_divu(%lhs: memref<7x14x21xi32>, %rhs: memref<7x14x21xi32>,
+                          %out: memref<7x14x21xi32>) {
+  linalg.div_unsigned ins(%lhs, %rhs : memref<7x14x21xi32>, memref<7x14x21xi32>)
+             outs(%out : memref<7x14x21xi32>)
+  return
+}
+
+// CHECK-DAG: #[[MAP:.+]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
+
+// CHECK: func @generalize_divu
+// CHECK-SAME: (%[[LHS:.+]]: memref<7x14x21xi32>, %[[RHS:.+]]: memref<7x14x21xi32>,
+// CHECK-SAME:  %[[OUT:.+]]: memref<7x14x21xi32>)
+
+// CHECK: linalg.generic
+// CHECK-SAME: indexing_maps = [#[[MAP]], #[[MAP]], #[[MAP]]]
+// CHECK-SAME: iterator_types = ["parallel", "parallel", "parallel"]}
+// CHECK-SAME:  ins(%[[LHS]], %[[RHS]] : memref<7x14x21xi32>, memref<7x14x21xi32>)
+// CHECK-SAME: outs(%[[OUT]] : memref<7x14x21xi32>)
+
+// CHECK:         ^{{.+}}(%[[BBARG0:.+]]: i32, %[[BBARG1:.+]]: i32, %[[BBARG2:.+]]: i32)
+// CHECK-NEXT:      %[[SUM:.+]] = arith.divui %[[BBARG0]], %[[BBARG1]] : i32
+// CHECK-NEXT:      linalg.yield %[[SUM]] : i32

mlir/test/Dialect/Linalg/named-ops-fail.mlir

 // RUN: not mlir-opt -split-input-file -verify-diagnostics %s 2>&1 | FileCheck %s
 
 func.func @add_type_cast(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf16>, %arg2: memref<4x8x16xf32>) {
   // CHECK: op requires the same type for all operands and results
   linalg.add ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf16>) outs(%arg2: memref<4x8x16xf32>)
   return
 }
 
 // -----
 
 func.func @add_broadcast(%arg0: memref<8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
   // CHECK: op expected operand rank (2) to match the result rank of indexing_map #0 (3)
   linalg.add ins(%arg0, %arg1 : memref<8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
   return
 }
 
+// -----
+
+func.func @sub_type_cast(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf16>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op requires the same type for all operands and results
+  linalg.sub ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf16>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @sub_broadcast(%arg0: memref<8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op expected operand rank (2) to match the result rank of indexing_map #0 (3)
+  linalg.sub ins(%arg0, %arg1 : memref<8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @mul_type_cast(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf16>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op requires the same type for all operands and results
+  linalg.mul ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf16>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @mul_broadcast(%arg0: memref<8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op expected operand rank (2) to match the result rank of indexing_map #0 (3)
+  linalg.mul ins(%arg0, %arg1 : memref<8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @div_type_cast(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf16>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op requires the same type for all operands and results
+  linalg.div ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf16>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @div_broadcast(%arg0: memref<8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op expected operand rank (2) to match the result rank of indexing_map #0 (3)
+  linalg.div ins(%arg0, %arg1 : memref<8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @divu_type_cast(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf16>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op requires the same type for all operands and results
+  linalg.div_unsigned ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf16>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+func.func @divu_broadcast(%arg0: memref<8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: op expected operand rank (2) to match the result rank of indexing_map #0 (3)
+  linalg.div_unsigned ins(%arg0, %arg1 : memref<8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}

mlir/test/Dialect/Linalg/named-ops.mlir

 func.func @add_tensor(%arg0: tensor<4x8x16xf32>, %arg1: tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
   %0 = tensor.empty() : tensor<4x8x16xf32>
   // CHECK: linalg.add
   // CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<4x8x16xf32>, tensor<4x8x16xf32>)
   // CHECK-SAME: outs(%{{.+}} : tensor<4x8x16xf32>)
   %1 = linalg.add ins(%arg0, %arg1 : tensor<4x8x16xf32>, tensor<4x8x16xf32>) outs(%0: tensor<4x8x16xf32>) -> tensor<4x8x16xf32>
   return %1 : tensor<4x8x16xf32>
 }
+
+// -----
+
+// CHECK-LABEL: func @sub_dynamic
+func.func @sub_dynamic(%arg0: memref<?x?x?xf32>, %arg1: memref<?x?x?xf32>, %arg2: memref<?x?x?xf32>) {
+  // CHECK: linalg.sub
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
+  linalg.sub ins(%arg0, %arg1 : memref<?x?x?xf32>, memref<?x?x?xf32>) outs(%arg2: memref<?x?x?xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @sub_static
+func.func @sub_static(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: linalg.sub
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<4x8x16xf32>, memref<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<4x8x16xf32>)
+  linalg.sub ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @sub_tensor
+func.func @sub_tensor(%arg0: tensor<4x8x16xf32>, %arg1: tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
+  %0 = tensor.empty() : tensor<4x8x16xf32>
+  // CHECK: linalg.sub
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<4x8x16xf32>, tensor<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : tensor<4x8x16xf32>)
+  %1 = linalg.sub ins(%arg0, %arg1 : tensor<4x8x16xf32>, tensor<4x8x16xf32>) outs(%0: tensor<4x8x16xf32>) -> tensor<4x8x16xf32>
+  return %1 : tensor<4x8x16xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func @mul_dynamic
+func.func @mul_dynamic(%arg0: memref<?x?x?xf32>, %arg1: memref<?x?x?xf32>, %arg2: memref<?x?x?xf32>) {
+  // CHECK: linalg.mul
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
+  linalg.mul ins(%arg0, %arg1 : memref<?x?x?xf32>, memref<?x?x?xf32>) outs(%arg2: memref<?x?x?xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @mul_static
+func.func @mul_static(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: linalg.mul
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<4x8x16xf32>, memref<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<4x8x16xf32>)
+  linalg.mul ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @mul_tensor
+func.func @mul_tensor(%arg0: tensor<4x8x16xf32>, %arg1: tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
+  %0 = tensor.empty() : tensor<4x8x16xf32>
+  // CHECK: linalg.mul
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<4x8x16xf32>, tensor<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : tensor<4x8x16xf32>)
+  %1 = linalg.mul ins(%arg0, %arg1 : tensor<4x8x16xf32>, tensor<4x8x16xf32>) outs(%0: tensor<4x8x16xf32>) -> tensor<4x8x16xf32>
+  return %1 : tensor<4x8x16xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func @div_dynamic
+func.func @div_dynamic(%arg0: memref<?x?x?xf32>, %arg1: memref<?x?x?xf32>, %arg2: memref<?x?x?xf32>) {
+  // CHECK: linalg.div
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
+  linalg.div ins(%arg0, %arg1 : memref<?x?x?xf32>, memref<?x?x?xf32>) outs(%arg2: memref<?x?x?xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @div_static
+func.func @div_static(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: linalg.div
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<4x8x16xf32>, memref<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<4x8x16xf32>)
+  linalg.div ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @div_tensor
+func.func @div_tensor(%arg0: tensor<4x8x16xf32>, %arg1: tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
+  %0 = tensor.empty() : tensor<4x8x16xf32>
+  // CHECK: linalg.div
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<4x8x16xf32>, tensor<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : tensor<4x8x16xf32>)
+  %1 = linalg.div ins(%arg0, %arg1 : tensor<4x8x16xf32>, tensor<4x8x16xf32>) outs(%0: tensor<4x8x16xf32>) -> tensor<4x8x16xf32>
+  return %1 : tensor<4x8x16xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func @div_unsigned_dynamic
+func.func @div_unsigned_dynamic(%arg0: memref<?x?x?xf32>, %arg1: memref<?x?x?xf32>, %arg2: memref<?x?x?xf32>) {
+  // CHECK: linalg.div_unsigned
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<?x?x?xf32>, memref<?x?x?xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<?x?x?xf32>)
+  linalg.div_unsigned ins(%arg0, %arg1 : memref<?x?x?xf32>, memref<?x?x?xf32>) outs(%arg2: memref<?x?x?xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @div_unsigned_static
+func.func @div_unsigned_static(%arg0: memref<4x8x16xf32>, %arg1: memref<4x8x16xf32>, %arg2: memref<4x8x16xf32>) {
+  // CHECK: linalg.div_unsigned
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : memref<4x8x16xf32>, memref<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : memref<4x8x16xf32>)
+  linalg.div_unsigned ins(%arg0, %arg1 : memref<4x8x16xf32>, memref<4x8x16xf32>) outs(%arg2: memref<4x8x16xf32>)
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @div_unsigned_tensor
+func.func @div_unsigned_tensor(%arg0: tensor<4x8x16xf32>, %arg1: tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
+  %0 = tensor.empty() : tensor<4x8x16xf32>
+  // CHECK: linalg.div_unsigned
+  // CHECK-SAME: ins(%{{.+}}, %{{.+}} : tensor<4x8x16xf32>, tensor<4x8x16xf32>)
+  // CHECK-SAME: outs(%{{.+}} : tensor<4x8x16xf32>)
+  %1 = linalg.div_unsigned ins(%arg0, %arg1 : tensor<4x8x16xf32>, tensor<4x8x16xf32>) outs(%0: tensor<4x8x16xf32>) -> tensor<4x8x16xf32>
+  return %1 : tensor<4x8x16xf32>
+}