%4 = "tf.If"(%3, %3, %arg0, %cst) {_tpu_replicate = "cluster", _xla_outside_compilation = "0", device = "", else_branch = @if_else_branch, is_stateless = false, then_branch = @if_then_branch} : (tensor<*xi1>, tensor<*xi1>, tensor<*xi64>, tensor<i64>) -> tensor<*xi1>
    func.return %4 : tensor<*xi1>
  }

// CHECK: func private @[[FUNCTIONALIZE_FUNC]]
// CHECK: "tf.If"
// CHECK-SAME:  else_branch = @[[ELSE_FUNC:[A-Za-z0-9_]*]]
// CHECK-SAME:  then_branch = @[[THEN_FUNC:[A-Za-z0-9_]*]]

// We expect the _tf.Add in the else func and the _tf.Mul in the then func

// CHECK: func private @[[ELSE_FUNC]]
// CHECK: "tf.Add"
// CHECK: func private @[[THEN_FUNC]]

# Tests V1 control flow is functionalized.

# CHECK-NOT: tf_executor.Switch
# CHECK-NOT: tf_executor.Merge
# CHECK: "tf.If"
# CHECK-SAME: else_branch = @"key/[[else:[a-zA-Z_0-9]+]]"
# CHECK-SAME: then_branch = @"key/[[then:[a-zA-Z_0-9]+]]"

# CHECK: func private @"key/[[else]]"(
# CHECK-SAME: tf._original_func_name
# CHECK: func private @"key/[[then]]"(
# CHECK-SAME: tf._original_func_name


def Test():

  // CHECK: [[cond:%.*]] = "tf.Const"()
  %cond = "tf.Const"() {device = "/CPU:0", value = dense<true> : tensor<i1>} : () -> tensor<i1>
  // CHECK: "tf.If"([[cond]], [[handle]])
  %x = "tf.If"(%cond, %handle) {then_branch = @some_other_func, else_branch = @some_other_func, is_stateless = false} : (tensor<i1>, tensor<!tf_type.resource<tensor<i32>>>) -> tensor<i32>
  %r = "tf.AddV2"(%arg, %x) {device = "/CPU:0"} : (tensor<i32>, tensor<i32>) -> tensor<i32>

  %18 = "tf.Identity"(%17) {device = ""} : (tensor<i1>) -> tensor<i1>

// CHECK: [[z:%.*]] = mlrt.cond [[cond]] @then @else
// CHECK: return [[z]]
func.func @while_cond_if(%cond: tensor<i1>, %x: tensor<i1>, %y: tensor<i1>, %z: tensor<i32>) -> (tensor<i1>) {
  %r = "tf.If"(%cond, %x, %y, %z) {then_branch = @then, else_branch = @else, is_stateless = true} : (tensor<i1>, tensor<i1>, tensor<i1>, tensor<i32>) -> tensor<i1>
  return %r : tensor<i1>
}

// CHECK-LABEL: func @while_body_if

    } else if (auto if_op = llvm::dyn_cast<TF::IfOp>(use.getOwner())) {
      auto then_branch = if_op.then_function();
      auto else_branch = if_op.else_function();
      assert(then_branch && else_branch);
      auto type_from_then = GetElementTypeFromAccess(
          then_branch.getArgument(use.getOperandNumber() - 1), module,
          infer_from_op);

        : () -> !tf_res
      "tf.If"(%arg1, %id0, %var_handle) {
          then_branch = @if_then_and_else,
          else_branch = @if_then_and_else, is_stateless = false}
        : (tensor<i1>, !tf_res,
           !tf_res) -> ()
      "tf.If"(%arg1, %var_handle, %id0) {
      // expected-error@above {{Conflicting device assignment for resource}}
          then_branch = @if_then_and_else,
          else_branch = @if_then_and_else,

                         llvm::StringMap<PartitionedCallTensorArrayOpsInfo>*
                             decomposed_partitioned_call_callees) {
  auto then_branch = if_op.then_function();
  auto else_branch = if_op.else_function();
  auto grads = AccessedGradients({then_branch, else_branch}, module);
  auto ta_arg_buffer_type = [&](int64_t index) -> Type {
    auto it = stats->find(if_op.getOperand(index + 1));

func.func @main(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  %0 = "tfl.less"(%arg0, %arg1) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xi1>
  %1 = "tf.If"(%0, %arg0, %arg1) {else_branch = @cond_false, then_branch = @cond_true, is_stateless = false} : (tensor<1xi1>, tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  func.return %1 : tensor<1xf32>
}

func.func @cond_true(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>) -> tensor<*xf32> {