// deferred.

// CHECK-LABEL: add_with_activation_transpose_rank_two
func.func @add_with_activation_transpose_rank_two(%arg0: tensor<1x2xf32>) -> tensor<2x1xf32> {
  %0 = stablehlo.constant dense<2.000000e+00> : tensor<2x1xf32>
  %1 = stablehlo.transpose %arg0, dims = [1, 0] : (tensor<1x2xf32>) -> tensor<2x1xf32>
  %2 = stablehlo.add %1, %0 : tensor<2x1xf32>
  return %2 : tensor<2x1xf32>
}

}

// CHECK-LABEL: QuantizeConcat
func.func @QuantizeConcat(tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<2x2x!quant.uniform<u8:f32, 1.000000e-01:128>> {
^bb0(%arg0: tensor<1x2xf32>, %arg1: tensor<1x2xf32>):
  %0 = "tfl.concatenation"(%arg0, %arg1) {axis = 0 : i32, fused_activation_function = "NONE"} : (tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<2x2xf32>

  // CHECK-NEXT:     %[[GROUP:.*]] = "tf.Const"()
  // CHECK-SAME:       [0, 1, 2, 3]
  // CHECK-NEXT:     %[[REDUCED:.*]] = "tf.XlaAllReduce"(%[[ID]], %[[GROUP]]) <{mode = "CrossReplica", reduce_op = "Add"}> : (tensor<f32>, tensor<1x4xi32>) -> tensor<f32>
  // CHECK-NEXT:     "tf.OpA"(%[[REDUCED]]) : (tensor<f32>) -> ()
  tf_device.replicate {n = 4 : i32} {
    "tf_device.cluster"() ({
      "tf.OpA"(%arg0) : (tensor<f32>) -> ()
      tf_device.return

  %0 = "tf_device.cluster_func"(%arg0, %arg1) {func = @_func, use_spmd_for_xla_partitioning = true, use_tpu = true, num_cores_per_replica = 2 : i64} : (tensor<2x4xf32>, tensor<4x2xf32>) -> tensor<2x2xf32>
  %1:2 = "tf.TPUPartitionedOutputV2"(%0) {device = "", partition_dims = [2, 1]} : (tensor<2x2xf32>) -> (tensor<1x2xf32>, tensor<1x2xf32>)

// CHECK-NEXT:  %[[argmax:.*]] = "tfl.arg_max"(%[[softmax]], %[[cst]]) : (tensor<128x16xf32>, tensor<i32>) -> tensor<128xi32>
  %3 = "tfl.arg_max"(%2, %cst) : (tensor<128x16xf32>, tensor<i32>) -> tensor<128xi32>
// CHECK-NEXT:  return %[[softmax]], %[[argmax]] : tensor<128x16xf32>, tensor<128xi32>
  func.return %2, %3 : tensor<128x16xf32>, tensor<128xi32>
}

// CHECK-LABEL: PruneUnusedLstm

// CHECK:}

func.func @reshape(%arg0 : tensor<1x128xi32>) -> tensor<4x32x1xi32>{
  %0 = "vhlo.reshape_v1"(%arg0) : (tensor<1x128xi32>) -> tensor<4x32x1xi32>
  func.return %0 : tensor<4x32x1xi32>
}

//CHECK:func.func private @reshape(%arg0: tensor<1x128xi32>) -> tensor<4x32x1xi32> {
//CHECK-NEXT: %0 = "vhlo.reshape_v1"(%arg0) : (tensor<1x128xi32>) -> tensor<4x32x1xi32>
//CHECK-NEXT: return %0 : tensor<4x32x1xi32>

  %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>

  func.return %3 : tensor<10xi32>
// CHECK: %[[CAST_0:.*]] = "tfr.cast"(%arg0) : (tensor<10x!quant.uniform<i8:f32, 0.0039133410900831223:-128>>) -> !tfr.tensor
// CHECK: %[[CAST_1:.*]] = "tfr.cast"(%[[CAST_0]]) : (!tfr.tensor) -> tensor<10xi8>
// CHECK: %[[CAST_2:.*]] = "tf.Cast"(%[[CAST_1]]) <{Truncate = false}> : (tensor<10xi8>) -> tensor<10xi32>
// CHECK: return %[[CAST_2]] : tensor<10xi32>
}

// `tfl.concatenation`.

func.func @concatenate_float(%arg0: tensor<3x2xf32>, %arg1: tensor<1x2xf32>) -> tensor<4x2xf32> {
  %0 = "stablehlo.concatenate"(%arg0, %arg1) {dimension = 0 : i64} : (tensor<3x2xf32>, tensor<1x2xf32>) -> tensor<4x2xf32>
  return %0 : tensor<4x2xf32>
}
// CHECK-LABEL: concatenate_float
// CHECK-NOT: tfl.concatenation
// CHECK: stablehlo.concatenate

// -----

  func.func @dynamic_update_slice(%arg0: tensor<3x4xi32>, %arg1: tensor<2x2xi32>, %arg2: tensor<2xi32>) -> tensor<3x4xi32> {

    // CHECK: %[[SLICE0:.*]] = "mhlo.slice"(%[[ARG2]])
    // CHECK-DAG-SAME: start_indices = dense<0> : tensor<1xi64>
    // CHECK-DAG-SAME: limit_indices = dense<1> : tensor<1xi64>
    // CHECK-DAG-SAME: strides = dense<1> : tensor<1xi64>