func.return %0#0 : tensor<2xi32>
}

// -----

func.func @unpack(%arg0: tensor<*xi32>) -> (tensor<*xi32>, tensor<*xi32>) {
  %0:2 = "tfl.unpack"(%arg0) {axis = 1 : i32, num = 2 : i32} : (tensor<*xi32>) -> (tensor<*xi32>, tensor<*xi32>)
  func.return %0#0, %0#1 : tensor<*xi32>, tensor<*xi32>
}

// -----

func.func @L2NormalizePattern(%arg0: tensor<2xf32>) -> tensor<2xf32> {
  %cst = arith.constant dense<[0]> : tensor<1xi32>
  %0 = "tfl.square"(%arg0) : (tensor<2xf32>) -> tensor<2xf32>
  %1 = "tfl.sum"(%0, %cst) {keep_dims = false} : (tensor<2xf32>, tensor<1xi32>) -> tensor<f32>
  %2 = "tfl.rsqrt"(%1) : (tensor<f32>) -> tensor<f32>
  %3 = "tfl.mul"(%arg0, %2) {fused_activation_function = "NONE"} : (tensor<2xf32>, tensor<f32>) -> tensor<2xf32>

// -----

// CHECK-LABEL: func @floordiv_int
func.func @floordiv_int(%arg0: tensor<?xi32>, %arg1: tensor<?xi32>) -> tensor<?xi32> {
  // CHECK-DAG: [[DIV:%.+]] = chlo.broadcast_divide %arg0, %arg1 : (tensor<?xi32>, tensor<?xi32>) -> tensor<?xi32>
  // CHECK-DAG: [[MUL:%.+]] = chlo.broadcast_multiply [[DIV]], %arg1 : (tensor<?xi32>, tensor<?xi32>) -> tensor<?xi32>

  // CHECK-LABEL: func @main(%arg0: tensor<1xi32>, %arg1: tensor<1xi32>) -> tensor<1xi32>
  func.func @main(%arg0: tensor<1xi32>, %arg1: tensor<1xi32>) -> tensor<*xi32> {
    // CHECK: %[[RESULT:.*]] = "tf.AddV2"
    // CHECK-SAME: (tensor<1xi32>, tensor<1xi32>) -> tensor<1xi32>
    // CHECK: return %[[RESULT]] : tensor<1xi32>
    %0 = "tf.Cast"(%arg0) : (tensor<1xi32>) -> tensor<*xi32>
    %1 = "tf.Cast"(%arg1) : (tensor<1xi32>) -> tensor<*xi32>

func.func private @simple_callee() -> tensor<2xi32>  {
  %cst = "tf.Const"() { value = dense<2> : tensor<2xi32> } : () -> tensor<2xi32>
  func.return %cst : tensor<2xi32>
}

func.func private @simple_callee_with_noinline() -> tensor<2xi32> attributes {tf._noinline = true} {
  %cst = "tf.Const"() { value = dense<2> : tensor<2xi32> } : () -> tensor<2xi32>
  func.return %cst : tensor<2xi32>
}

// Test that simple TF operations can be inlined.

// CHECK-NEXT:      %6 = "tfl.pseudo_const"() <{value = dense<1> : tensor<i32>}> : () -> tensor<i32>
// CHECK-NEXT:      %7 = "tfl.unsorted_segment_prod"(%3, %4, %6) : (tensor<4xi32>, tensor<4xi32>, tensor<i32>) -> tensor<1xi32>
// CHECK-NEXT:      %8 = "tfl.unsorted_segment_prod"(%3, %5, %6) : (tensor<4xi32>, tensor<4xi32>, tensor<i32>) -> tensor<1xi32>

func.func @listPopBack(%arg0: tensor<!tf_type.variant<tensor<2xi32>>>, %arg1: tensor<1xi32>) -> (tensor<!tf_type.variant<tensor<2xi32>>>, tensor<2xi32>)  {
  %0, %1 = "tf.TensorListPopBack"(%arg0, %arg1) : (tensor<!tf_type.variant<tensor<2xi32>>>, tensor<1xi32>) -> (tensor<!tf_type.variant<tensor<2xi32>>>, tensor<2xi32>)

// CHECK:           [[VAL_23:%.*]] = "tf.Reshape"([[VAL_15]]#1, [[VAL_22]]) : (tensor<1x10xf32>, tensor<1xi32>) -> tensor<10xf32>
// CHECK-DAG:       [[VAL_24:%.*]] = "tf.Const"() <{value = dense<-1> : tensor<1xi32>}> : () -> tensor<1xi32>
// CHECK:           [[VAL_25:%.*]] = "tf.Reshape"([[VAL_15]]#2, [[VAL_24]]) : (tensor<1x10xf32>, tensor<1xi32>) -> tensor<10xf32>

      %17 = "tfl.reduce_prod"(%16, %cst_14) {keep_dims = true, tac.device = "CPU", tac.inference_type = "FLOAT"} : (tensor<2xi32>, tensor<1xi32>) -> tensor<1xi32>
      %18 = "tfl.reshape"(%arg1, %17) {tac.device = "DARWINN", tac.inference_type = "FLOAT"} : (tensor<?x?x!tf_type.string>, tensor<1xi32>) -> tensor<?x!tf_type.string>

// CHECK: %[[v5:.*]] = "tf.Gather"(%3, %[[cst_1]]) <{validate_indices = true}> : (tensor<3xi32>, tensor<2xi32>) -> tensor<2xi32>
// CHECK: %[[v6:.*]] = "tf.Gather"(%4, %[[cst_2]]) <{validate_indices = true}> : (tensor<4xi32>, tensor<1xi32>) -> tensor<1xi32>
// CHECK: %[[v7:.*]] = "tf.Concat"(%[[cst_3]], %5, %6) : (tensor<i32>, tensor<2xi32>, tensor<1xi32>) -> tensor<3xi32>