%6 = "tfl.reshape"(%5, %cst_1) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<200xf32>, tensor<2xi32>) -> tensor<2x100xf32>
    %7 = "tfl.quantize"(%6) {qtype = tensor<2x100x!quant.uniform<i8:f32, 2.000000e-01:-3>>, tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<2x100xf32>) -> tensor<2x100x!quant.uniform<i8:f32, 2.000000e-01:-3>>

  %pi_0 = "tf.TPUPartitionedInput"(%arg0, %arg1) {_XlaSharding = "123", partition_dim = 0 : i64} : (tensor<10x3xf32>, tensor<10x3xf32>) -> tensor<20x3xf32>
  %ri = "tf.TPUReplicatedInput"(%pi_0) : (tensor<20x3xf32>) -> tensor<20x3xf32>
  // CHECK: return [[RI]]
  func.return %ri : tensor<20x3xf32>
}

// -----

// CHECK-LABEL:func @partitioned_2d_resource

    : ( tensor<1x28x28xf32>,
        tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>,
        tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>,
        tensor<20xf32>, tensor<20xf32>, tensor<20xf32>,
        tensor<20xf32>, tensor<20xf32>, tensor<20xf32>, tensor<20xf32>,
        tensor<20x20xf32>, none,
        tensor<1x20xf32>, tensor<1x20xf32>,

  %0:2 = "tfl.fully_connected"(%arg0, %arg1, %cst) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<40x37xf32>, tensor<40x37xf32>, none) -> (tensor<40x40xf32>, tensor<40x40xf32>)
  func.return %0 : tensor<40x40xf32>

  %0:2 = "tfl.fully_connected"(%arg0, %arg1, %cst) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "SHUFFLED4x16INT8"} : (tensor<40x37xf32>, tensor<40x37xf32>, none) -> (tensor<40x40xf32>, tensor<40x40xf32>)
  func.return %0 : tensor<40x40xf32>

  // CHECK-LABEL: @get_next
  func.func private @get_next(%arg0: tensor<!tf_type.resource>) -> tensor<200x10xf32> {
    // CHECK: %0 = "tf.Iterator"
    // CHECK-SAME: shared_name = "foo_iterator"
    // CHECK: "tf.IteratorGetNext"(%0)
    %0 = "tf.IteratorGetNext"(%arg0) : (tensor<!tf_type.resource>) -> tensor<200x10xf32>
    return %0 : tensor<200x10xf32>
  }

  // CHECK-LABEL: @main
  func.func @main()

// CHECK-LABEL: float_matmul
func.func @float_matmul(
  %arg0: tensor<1x10xf32>, %arg1: tensor<10x10xf32>) -> (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) {
  %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32>
  %0 = "tf.MatMul"(%arg0, %arg1) {
    transpose_a = false, transpose_b = false
  } : (tensor<1x10xf32>, tensor<10x10xf32>) -> tensor<*xf32>

  }

  // CHECK-LABEL: pack
  func.func @pack(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<2x1xf32> {
    %0 = "tfl.pack"(%arg0, %arg1) {axis = 0 : i32, values_count = 2 : i32} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
    func.return %0 : tensor<2x1xf32>
    // CHECK: %[[VAL_0:.*]] = arith.constant dense<[2, 1]> : tensor<2xi32>

  // CHECK-LABEL: func @parallel_execute_with_tiled_input
  // CHECK-SAME: (%[[ARG_0:[a-z0-9]*]]: tensor<128x10xf32>, %[[ARG_1:[a-z0-9]*]]: tensor<128x10xf32>, %[[ARG_2:[a-z0-9]*]]: tensor<*xi32>, %[[ARG_3:[a-z0-9]*]]: tensor<*xi32>)
  func.func @parallel_execute_with_tiled_input(%arg0: tensor<128x10xf32>, %arg1: tensor<128x10xf32>, %arg2: tensor<*xi32>, %arg3: tensor<*xi32>) -> (tensor<*xi32>, tensor<*xi1>) {
    // CHECK: tf_device.replicate

  func.func @matmul(%arg0: tensor<2x12xf32>) -> (tensor<*xf32>) {
    %cst_0 = "tf.Const"() {value = dense<0.000000e+00> : tensor<12x2xf32>} : () -> tensor<12x2xf32>
    %1 = "tf.PartitionedCall"(%arg0, %cst_0) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_matmul_fn_1} : (tensor<2x12xf32>, tensor<12x2xf32>) -> tensor<*xf32>
    func.return %1: tensor<*xf32>
  }