func.return
}

// -----

func.func @parallel_execute_regions_with_data_results(%arg0: tensor<i32>) -> tensor<i32> {
  // expected-error @+1 {{'func' attribute refers to an undefined function: undefined_func}}
  %0 = "tf_device.cluster_func"(%arg0) {func = @undefined_func} : (tensor<i32>) -> tensor<i32>
  func.return %0 : tensor<i32>

  %1 = "tf.Const"() { value = dense<42.0> : tensor<f32> } : () -> tensor<f32>
  func.return %0, %1 : tensor<i32>, tensor<f32>
}

// CHECK: @cond(%arg0{{.*}}, %arg1{{.*}})
func.func private @cond(%arg0: tensor<i32>, %arg1: tensor<f32>) -> tensor<i1> {
  %const = "tf.Const"() { value = dense<1000> : tensor<i32> } : () -> tensor<i32>
  %result = "tf.Less"(%const, %arg0) : (tensor<i32>, tensor<i32>) -> tensor<i1>

  // CHECK-DAG: %0 = mhlo.constant dense<1> : tensor<i32>
  // CHECK: %1 = "mhlo.broadcast_in_dim"(%0) <{broadcast_dimensions = dense<> : tensor<0xi64>}> : (tensor<i32>) -> tensor<1x750xi32>
  // CHECK: return %1 : tensor<1x750xi32>
}

// CHECK-LABEL: @splat_scalar_no_change
func.func @splat_scalar_no_change() -> (tensor<f32>, tensor<i32>) {
  // CHECK-NOT: mhlo.broadcast_in_dim

  // CHECK-SAME: tfl.schema_version = 3 : i32

  // CHECK:          %{{.*}} = "tfl.pseudo_const"() <{value = dense<{{\[\[1, 2\], \[3, 4\], \[5, 6\]\]}}> : tensor<3x2xi32>}>
  // CHECK-NEXT:     [[SUB:%.*]] = tfl.sub %{{.*}}, %{{.*}} {fused_activation_function = "RELU6"} : tensor<3x2xi32>
  // CHECK-NEXT:     [[SCALAR:%.*]] = "tfl.pseudo_const"() <{value = dense<10> : tensor<i32>}> : () -> tensor<i32>

//CHECK-DAG:    %[[c1:.+]] = stablehlo.constant dense<1> : tensor<i32>
//CHECK-DAG:    %[[c2:.+]] = stablehlo.constant dense<0> : tensor<i32>
//CHECK-NEXT:    %[[c3:.+]] = stablehlo.dynamic_update_slice %arg0, %[[c0]], %[[c1]], %[[c2]], %[[c2]] : (tensor<2x1x2xf32>, tensor<1x1x2xf32>, tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<2x1x2xf32>
//CHECK-NEXT:    return %[[c3:.+]] : tensor<2x1x2xf32>
//CHECK-NEXT:  }

    tf_executor.graph {
      %c, %control0 = tf_executor.island wraps "tf.Const"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
      %a, %control1 = tf_executor.island wraps "tf.A"() {_class = ["loc:@class"]} : () -> (tensor<2xf32>)
      %s:2, %control2 = tf_executor.island wraps "tf.Split"(%c, %a) {num_split = 2 : i32} : (tensor<i32>, tensor<2xf32>) -> (tensor<1xf32>, tensor<1xf32>)
      tf_executor.fetch
    }
    func.return
  }

}

// -----
// CHECK-LABEL: func @test_ifrt_call
func.func @test_ifrt_call(%arg0: tensor<?xf32>, %arg1: tensor<?xf32>) {
  %result = "tf.IfrtCall"(%arg0, %arg1) <{program_id = 1234 : i64, variable_arg_indices = [0 : i32, 1 : i32], variable_names = ["a", "b"]}> : (tensor<?xf32>, tensor<?xf32>) -> (tensor<1x1xf32>)
  func.return
}

// -----
func.func @test_ifrt_call_fail_unsorted_variable_arg_indices(%arg0: tensor<?xf32>, %arg1: tensor<?xf32>) {

  func.return %5 : tensor<!tf_type.variant>
}

func.func @"__inference_Dataset_map_<lambda>_80"(%arg0: tensor<*xi32>) -> tensor<*xi32> {
  %0 = "tf.Const"() {value = dense<2> : tensor<i32>} : () -> tensor<i32>
  %1 = "tf.Mul"(%arg0, %0) {device = ""} : (tensor<*xi32>, tensor<i32>) -> tensor<*xi32>
  %2 = "tf.Identity"(%1) {device = ""} : (tensor<*xi32>) -> tensor<*xi32>
  func.return %2 : tensor<*xi32>
}

func.func @main(%arg0: tensor<1x1024xf32>) -> tensor<1x3xf32> {
  %0 = "tf.Const"() <{value = dense<0.5> : tensor<1024x3xf32>}> : () -> tensor<1024x3xf32>
  %1:4 = "tf.CustomAggregator"(%arg0) <{id = "1", calibration_method = 1 : i32, num_bins = 0 : i32, max_percentile = 0.000000e+00 : f32, min_percentile = 0.000000e+00 : f32}> {min = 7.547870e-07 : f32, max = 0.999992311 : f32} : (tensor<1x1024xf32>) -> (tensor<1x1024xf32>, tensor<f32>, tensor<f32>, tensor<*xi64>)

  %2 = "tfl.pseudo_qconst"() {qtype = tensor<32x!quant.uniform<i32:f32, 1.7052092479439231E-4>>, value = dense<0> : tensor<32xi32>} : () -> tensor<32x!quant.uniform<i32:f32, 1.7052092479439231E-4>>