%3 = "tfl.pack"(%1, %2) {axis = 0 : i32, per_device_costs = {CPU = 2.0 : f32, GPU = -1.0 : f32}, values_count = 2 : i32, tac.device = "CPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
  func.return %3 : tensor<2x1xf32>
})";
  const std::string kExpectedFB = CreateRuntimeMetadata();
  mlir::DialectRegistry registry;
  registry.insert<mlir::TFL::TensorFlowLiteDialect, mlir::arith::ArithDialect,

  func.return
}

func.func private @write_chain_func(%arg0: tensor<1xf32>) -> (tensor<1xf32>) {
  %cst = "tf.Const"() <{value = dense<[[0, 1]]> : tensor<1x2xi32>}> : () -> tensor<1x2xi32>
  %0 = "tf.XlaAllReduce"(%arg0, %cst) <{mode = "CrossReplica", reduce_op = "Add"}> : (tensor<1xf32>, tensor<1x2xi32>) -> tensor<1xf32>
  return %0 : tensor<1xf32>

  %1 = "tf.ReadVariableOp"(%0) {device = "/CPU:0"} : (tensor<!tf_type.resource<tensor<2x1xf32>>>) -> tensor<2x1xf32>
  %2:2 = "tf.Split"(%cst, %1) {device = "/CPU:0"} : (tensor<i32>,  tensor<2x1xf32>) -> (tensor<1x1xf32>, tensor<1x1xf32>)
  %3 = "tf.TPUExecute"(%arg0, %2#0, %program#0) {_producer_name = "UNKNOWN", device = "/TPU:0"} : (tensor<1x1xf32>, tensor<1x1xf32>, tensor<3x!tf_type.string>) -> tensor<1x1xf32>

  %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>) {
  // expected-error@below {{variable_arg_indices must be sorted in ascending order}}

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

// CHECK-LABEL: mirror_pad
func.func @mirror_pad(%arg0: tensor<2x3xcomplex<f64>>) -> tensor<4x7xcomplex<f64>> {
  %0 = mhlo.constant dense<[[1, 1], [2, 2]]> : tensor<2x2xi32>
  // NO_FALLBACK: tf.MirrorPad

func.func @my_cast(%arg0: tensor<2x3xf32>) -> tensor<2x3xi32> {
  %0 = "tf.MyCast"(%arg0) {Tout=i32} : (tensor<2x3xf32>) -> tensor<2x3xi32>
  func.return %0 : tensor<2x3xi32>

// CHECK-NEXT: %[[CAST:.*]] = "tf.RiscCast"(%arg0) {Tout = i32} : (tensor<2x3xf32>) -> tensor<*xi32>
// CHECK-NEXT: %[[ES:.*]] = "tf.EnsureShape"(%[[CAST]]) <{shape = #tf_type.shape<2x3>}> : (tensor<*xi32>) -> tensor<2x3xi32>
// CHECK-NEXT: return %[[ES]] : tensor<2x3xi32>
}

// CHECK-DAG: %[[padding_rank_2:.*]] = "tf.Reshape"(%[[padding_rank_1]], {{.*}}) : (tensor<8xi32>, tensor<2xi64>) -> tensor<4x2xi32>
// CHECK-DAG: %[[input_padded:.*]] = "tf.PadV2"(%{{.*}}, %[[padding_rank_2]], {{.*}}) : (tensor<?x?x?x3xi8>, tensor<4x2xi32>, tensor<i8>) -> tensor<?x?x?x3xi8>

  %one = "glue.constant"() { value = 1: i32 } : () -> i32
  %done = "glue.compare" (%one, %one) { predicate = #glue<"compare LTE"> } : (i32, i32) -> i1
  %2 = mhlo.constant dense<[[1.1]]> : tensor<1x1xf32>
  %3 = mhlo.multiply %2, %2 : tensor<1x1xf32>
  %cst = "tf.Const"() {value = dense<0.0> : tensor<f32>} : () -> tensor<f32>
  %0 = "tf.AddV2"(%arg0, %cst) {device = "/device:CPU:0"} : (tensor<f32>, tensor<f32>) -> tensor<f32>

  %8 = "tf.MatMul"(%2, %7) {device = "/job:localhost/replica:0/task:0/device:CPU:0", transpose_a = false, transpose_b = false} : (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<3x3xf32>
  %9 = "tf.MatrixDeterminant"(%8) {T = f32, device = "/job:localhost/replica:0/task:0/device:CPU:0"} : (tensor<3x3xf32>) -> tensor<f32>

  %cst = "tf.Const"() {value = dense<0> : tensor<1x2xi32>} : () -> tensor<1x2xi32>
  %cst_0 = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
  %cst_1 = "tf.Const"() {value = dense<2> : tensor<1xi32>} : () -> tensor<1xi32>
  %cst_2 = "tf.Const"() {value = dense<4> : tensor<1x2xi32>} : () -> tensor<1x2xi32>
  %0 = "tf.SpaceToBatchND"(%arg0, %cst_1, %cst_2) : (tensor<1x128x3xf32>, tensor<1xi32>, tensor<1x2xi32>) -> tensor<2x68x3xf32>