}

func.func @FusingaddRelu(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  %0 = "tf.Add"(%arg0, %arg1) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %1 = "tf.Add"(%arg0, %0) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %2 = "tf.Relu"(%1) : (tensor<1xf32>) -> tensor<1xf32>
  %3 = "tf.Relu"(%arg0) : (tensor<1xf32>) -> tensor<1xf32>
  %4 = "tf.Add"(%3, %2) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>

module {
  // CHECK-LABEL: main
  func.func @main(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>) -> tensor<4xf32> {
    %0 = "tfl.squared_difference"(%arg0, %arg1) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
    func.return %0 : tensor<4xf32>
    // CHECK:  [[VAL_0:%.*]] = tfl.sub %arg0, %arg1 {fused_activation_function = "NONE"} : tensor<4xf32>

  %1:4 = "tfl.split"(%cst, %0) {num_splits = 4 : i32} : (tensor<i32>, tensor<4xf32>) -> (tensor<2xf32>, tensor<2xf32>,tensor<2xf32>, tensor<2xf32>)
  %2 = "tfl.quantize"(%1#0) {qtype = tensor<2x!quant.uniform<u8:f32, 1.0>>} : (tensor<2xf32>) -> tensor<2x!quant.uniform<u8:f32, 1.0>>
  %3 = "tfl.quantize"(%1#1) {qtype = tensor<2x!quant.uniform<u8:f32, 1.0>>} : (tensor<2xf32>) -> tensor<2x!quant.uniform<u8:f32, 1.0>>

    %1 = "tf.BiasAdd"(%0, %cst_0) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<3xf32>) -> tensor<*xf32>
    %2 = "tf.PartitionedCall"(%arg0, %cst_2) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_depthwise_conv2d_fn_1} : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<*xf32>
    func.return %1, %2: tensor<*xf32>, tensor<*xf32>

    %0 = "tf.Sub"(%cst_0, %cst_1) : (tensor<8xf32>, tensor<8xf32>) -> tensor<8xf32>
    return %0 : tensor<8xf32>
  }
// CHECK: @serving_default
// CHECK-DAG: %[[VAR_HANDLE_2:.*]] = "tf.VarHandleOp"() {{.*shared_name = "const_0".*}} : () -> tensor<!tf_type.resource<tensor<8xf32>>>
// CHECK-DAG: %[[READ_VAR_0:.*]] = "tf.ReadVariableOp"(%[[VAR_HANDLE_2]]) : (tensor<!tf_type.resource<tensor<8xf32>>>) -> tensor<8xf32>

  %2 = "tf.Relu6"(%1) {device = ""} : (tensor<*xf32>) -> tensor<*xf32>
  func.return %2 : tensor<*xf32>
}

// CHECK-LABEL: func @conv_with_non_constant_filter
// CHECK-DAG: %[[CONST_0:.*]] = "tf.Const"() <{value = dense<0.000000e+00> : tensor<2xf32>}> : () -> tensor<2xf32>
// CHECK: %[[PARTITIONEDCALL_0:.*]] = "tf.PartitionedCall"(%arg0, %arg1, %[[CONST_0]])
// CHECK-SAME: f = @composite_conv2d_with_bias_and_relu6_fn_1

// CHECK:           return %[[VAL_2]], %[[VAL_3]] : tensor<4xf32>, tensor<4xf32>
// CHECK:         }
func.func @broadcast_atan2(%arg0: tensor<1xf32>, %arg1: tensor<4xf32>) -> (tensor<4xf32>, tensor<4xf32>) {
  %0 = "mhlo.broadcast_in_dim"(%arg0) <{broadcast_dimensions = dense<[0]> : tensor<1xi64>}> : (tensor<1xf32>) -> tensor<4xf32>
  %1 = mhlo.atan2 %0, %arg1 : tensor<4xf32>
  %2 = mhlo.atan2 %arg1, %0 : tensor<4xf32>

    %val = "tf.ReadVariableOp"(%arg1) : (tensor<*x!tf_type.resource>) -> (tensor<0xf32>)
    %0 = "tf.AddV2"(%val, %arg0) : (tensor<0xf32>, tensor<*xf32>) -> tensor<*xf32>
    func.return %0 : tensor<*xf32>
  }

  // CHECK: func private @testIfElse(%arg0: tensor<*xf32>) -> tensor<*xf32>
  func.func private @testIfElse(%arg0:tensor<*xf32>, %arg1:tensor<*x!tf_type.resource>) -> tensor<*xf32> {

  %0 = "quantfork.stats"(%arg0) {
    layerStats = dense<[-1.28e-5, 1.27e-5]> : tensor<2xf32>
  } : (tensor<1x5x5x2xf32>) -> tensor<1x5x5x2xf32>
  %w = arith.constant dense<[[[[-1.0, 1.0]]], [[[1.0, 2.0]]], [[[-2.0, 1.0]]]]> : tensor<3x1x1x2xf32>
  %b = arith.constant dense<0.0> : tensor<3xf32>
  %b2 = arith.constant dense<[1.0e-2, 2.1473647e1, -2.1473647e2]> : tensor<3xf32>
  %conv = "tfl.conv_2d"(%0, %w, %b) {

  %cst_1 = "tf.Const"() {value = dense<[0.3, 0.4, 0.3, 0.4, 0.3, 0.4, 0.3, 0.4]> : tensor<8xf32>} : () -> tensor<8xf32>