// CHECK-DAG:       %[[VAL_3:.*]] = "tfl.pseudo_const"{{.*}}dense<2> : tensor<1xi32>
// CHECK-DAG:       %[[VAL_4:.*]] = "tfl.pseudo_const"{{.*}}dense<[2, 1]> : tensor<2xi32>
// CHECK:           %[[VAL_5:.*]] = "tfl.reshape"(%[[VAL_0]], %[[VAL_2]]) : (tensor<1xf32>, tensor<4xi32>) -> tensor<1x1x1x1xf32>
// CHECK:           %[[VAL_6:.*]] = "tfl.reshape"(%[[VAL_1]], %[[VAL_2]]) : (tensor<1xf32>, tensor<4xi32>) -> tensor<1x1x1x1xf32>

    %cst = "tf.Const"() <{value = dense<[1]> : tensor<1xi32>}> : () -> tensor<1xi32>
    %0 = "tf.VarHandleOp"() <{container = "", shared_name = "Variable"}> : () -> tensor<!tf_type.resource<tensor<2x3xbf16>>>
    %1 = "tf.ResourceGather"(%0, %cst) <{batch_dims = 0 : i64, validate_indices = true}> : (tensor<!tf_type.resource<tensor<2x3xbf16>>>, tensor<1xi32>) -> tensor<1x3xbf16>
    return %1: tensor<1x3xbf16>
  }

  func.return %1: tensor<10x20x30xcomplex<f32>>

// CHECK:  %[[cst:.*]] = arith.constant dense<30> : tensor<1xi32>
// CHECK:  %[[rff:.*]] = "tf.RFFT"(%arg0, %[[cst]]) : (tensor<10x20x30xf32>, tensor<1xi32>) -> tensor<10x20x30xcomplex<f32>>
}

// CHECK-LABEL: not_fold_cast_fft_to_rfft

  %2 = "tfl.add"(%arg0, %arg3) {tac.device = "GPU", fused_activation_function = "RELU6", tac.inference_type = "FLOAT"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %3 = "tfl.pack"(%1, %2) {tac.device = "CPU", tac.inference_type = "FLOAT", axis = 0 : i32, values_count = 2 : i32} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
  return %3 : tensor<2x1xf32>
}
```

  %size = "tf.Const"() {device = "/device:CPU:0", value = dense<1> : tensor<i32>} : () -> (tensor<i32>)
  %indices = "tf.Const"() {device = "/device:CPU:0", value = dense<[0]> : tensor<1xi32>} : () -> (tensor<1xi32>)
  // CHECK: tfrt_fallback_async.executeop key({{.*}}) cost({{.*}}) device("/job:localhost/replica:0/task:0/device:CPU:0") "tf.TensorArrayV3"

module {
  func.func @simpleTest(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>, %arg2: tensor<1xf32>, %arg3: tensor<1xf32>) -> tensor<2x1xf32> {
    %0 = func.call @func_0_GPU_FLOAT(%arg0, %arg1, %arg2) {tac.interface_name = "func_0"} : (tensor<1xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
    %1 = func.call @func_1_GPU_FLOAT(%arg0, %arg3) {tac.interface_name = "func_1"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>

  %3:2 = "tf.TPUReplicatedOutput"(%2#0) : (tensor<?xi32>) -> (tensor<?xi32>, tensor<?xi32>)
  %5:2 = "tf.TPUPartitionedOutputV2"(%3#0) {_XlaSharding = "", partition_dims = []} : (tensor<?xi32>) -> (tensor<?xi32>, tensor<?xi32>)
  %7:2 = "tf.TPUPartitionedOutputV2"(%3#1) {_XlaSharding = "", partition_dims = []} : (tensor<?xi32>) -> (tensor<?xi32>, tensor<?xi32>)

            %cmp = "tf.Less"(%read, %cst) : (tensor<32xf32>, tensor<32xf32>) -> tensor<32xi1>
            %dims = arith.constant dense<0> : tensor<1xi32>
            %reduce = "tf.All"(%cmp, %dims) {keep_dims = false} : (tensor<32xi1>, tensor<1xi32>) -> tensor<i1>
            "tf.Yield"(%reduce) : (tensor<i1>) -> ()
        }, {
          ^bb0(%barg0: tensor<i32>, %barg1: !tf_res, %barg2: !tf_res):

    %cst = "tf.Const"() {value = dense<-1> : tensor<i64>} : () -> tensor<i64>
    %cst_0 = "tf.Const"() {value = dense<[1, 0]> : tensor<2xi32>} : () -> tensor<2xi32>
    %cst_1 = "tf.Const"() {value = dense<1.000000e+00> : tensor<f32>} : () -> tensor<f32>
    %cst_2 = "tf.Const"() {value = dense<0.00235294132> : tensor<f32>} : () -> tensor<f32>

}

// CHECK-LABEL: QuantizeStridedSlice
func.func @QuantizeStridedSlice(tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<1x2x2x5xf32> {
^bb0(%arg0: tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>, %arg1: tensor<1xi32>, %arg2: tensor<1xi32>, %arg3: tensor<1xi32>):
  %0 = "tfl.dequantize"(%arg0) : (tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>) -> tensor<12x2x2x5xf32>