// CHECK-LABEL: @add_f32(%arg0: tensor<3x3xbf16>, %arg1: tensor<3x3xbf16>) -> tensor<3x3xbf16>
func.func @add_f32(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<3x3xf32> {
  // CHECK-NOT: f32
  // CHECK: stablehlo.add
  %0 = stablehlo.add %arg0, %arg1: (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<3x3xf32>
  return %0 : tensor<3x3xf32>
}

// -----

// CHECK-LABEL: @add_f64(%arg0: tensor<3x3xbf16>, %arg1: tensor<3x3xbf16>) -> tensor<3x3xbf16>

# CHECK:      func {{@[a-zA-Z_0-9]+}}(
# CHECK-SAME:   [[ARG0:%.*]]: tensor<3x1xf32> {tf_saved_model.index_path = ["x"]},
# CHECK-SAME:   [[ARG1:%.*]]: tensor<!tf_type.resource<tensor<1x3xf32>>> {tf_saved_model.bound_input = @[[VAR]]})
# CHECK-SAME:             -> (tensor<3x3xf32> {tf_saved_model.index_path = ["r"]})

  func.return
}

// CHECK-LABEL: func @basic
func.func @basic(
    %arg0: tensor<3x1xf32>,
    %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>>) -> (tensor<3x3xf32>) {
  %1 = "tf.ReadVariableOp"(%arg1) {_output_shapes = ["tfshape$dim { size: 1 } dim { size: 3 }"], device = "/device:CPU:0", dtype = f32} : (tensor<!tf_type.resource<tensor<1x3xf32>>>) -> tensor<1x3xf32>

  static constexpr char kMlirLegalOpStr[] = R"(
   func.func @infeed_dequeue_tuple_dynamic_error() -> (tensor<3x3xf32>, tensor<4x?xf32>) {
     %0:2 = "tf.InfeedDequeueTuple"() : () -> (tensor<3x3xf32>, tensor<4x?xf32>) func.return %0#0, %0#1 : tensor<3x3xf32>, tensor<4x?xf32>
   })";
  CellReader<int64_t> legalize_failure_count(
      "/tensorflow/core/tf2xla/v1/mlir_failed_xla_legalize_tf_pass_count");

# CHECK-LABEL:      func @key(
# CHECK-SAME:   %[[ARG0:.*]]: tensor<3x5xf32> {tf_saved_model.index_path = ["y"]}
# CHECK-SAME:   %[[ARG1:.*]]: tensor<5x3xf32> {tf_saved_model.index_path = ["x"]}
# CHECK-SAME:                  tensor<3x3xf32> {tf_saved_model.index_path = ["t"]}
# CHECK-SAME:                  tensor<5x5xf32> {tf_saved_model.index_path = ["s"]}
# CHECK-SAME: attributes {{.*}} tf_saved_model.exported_names = ["key"]

    func.func @main(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<6x3xf32> {
      %axis = "tf.Const"() { value = dense<0> : tensor<i64> } : () -> tensor<i64>
      %1 = "tf.ConcatV2"(%arg0, %arg1, %axis) : (tensor<3x3xf32>, tensor<3x3xf32>, tensor<i64>) -> tensor<6x3xf32>
      func.return %1 : tensor<6x3xf32>
    }
  })";
  CreateModule(kMlirModuleStr);

module {
  func.func public @main(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>, %arg2: tensor<1x100x32x4xf32>,
      %arg3: tensor<1x500x4x4xf32>, %arg4: tensor<1x500x4x4xf32>, %arg5: tensor<1x1x100x500xf32>, %arg6: tensor<f32>)
      -> tensor<1x100x32x4xf32> {

// CHECK: return

  func.func private @composite_dot_general_fn(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
      %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
      return %0 : tensor<1x3xf32>
  }
}

// -----

  %1 = "quantfork.stats"(%0) {bitsNum = 8 : i64, layerStats = dense<[-2.0, 2.0]> : tensor<2xf32>, narrowRange = false} : (tensor<2x3xf32>) -> tensor<2x3xf32>
  %2 = stablehlo.convert %1 : (tensor<2x3xf32>) -> (tensor<2x3xf32>)
  func.return %2 : tensor<2x3xf32>
}

// -----

// CHECK-LABEL: func @dot_redundant_stats

      %2 = "quantfork.stats"(%1) {layerStats = dense<[5.00000000e-6, 7.00000000e-1]> : tensor<2xf32>} : (tensor<1x3xf32>) -> tensor<1x3xf32>
      return %2 : tensor<1x3xf32>
    }

    func.func private @composite_dot_general_fn(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {