// 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>
}

    %recurrent_stats = "quantfork.stats"(%recurrent_input) {layerStats = dense<[-2.0, 1.0]> : tensor<2xf32>} : (tensor<1x20xf32>) -> tensor<1x20xf32>
    %cell_input = arith.constant dense<1.0> : tensor<1x20xf32>
    %cell_stats = "quantfork.stats"(%cell_input) {layerStats = dense<[-2.73090601, 7.94872093]> : tensor<2xf32>} : (tensor<1x20xf32>) -> tensor<1x20xf32>

// RUN: tf-quant-opt %s -split-input-file -quant-merge-duplicate-resource-ops | FileCheck %s

func.func @merge_duplicate_variable(%arg0: tensor<1x20xf32>, %arg1: tensor<!tf_type.string>) -> (tensor<20x4096xf32>) {
  %0 = tf_executor.graph {
    %outputs_5, %control_6 = tf_executor.island wraps "tf.Const"() {value = dense<""> : tensor<1x!tf_type.string>} : () -> tensor<1x!tf_type.string>

  %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>)

}

func.func @QDQsFollowedByTranspose(tensor<1x2xf32>) -> (tensor<2x1xf32>) {
^bb0(%arg0: tensor<1x2xf32>):
  %cst_0 = arith.constant dense<[1, 0]> : tensor<2xi32>
  %0 = "tfl.quantize"(%arg0){qtype = tensor<1x2x!quant.uniform<u8:f32, 1.0>>}: (tensor<1x2xf32>) -> (tensor<1x2x!quant.uniform<u8:f32, 1.0>>)
  %1 = "tfl.dequantize"(%0): (tensor<1x2x!quant.uniform<u8:f32, 1.0>>) -> (tensor<1x2xf32>)

    precision_config = #vhlo.array_v1<[#vhlo<precision_v1 DEFAULT>, #vhlo<precision_v1 DEFAULT>]>}> : (tensor<1x1x167xf32>, tensor<167x64xf32>) -> tensor<1x1x64xf32>
  return %0 : tensor<1x1x64xf32>
}

//CHECK:func.func private @dot_general(%arg0: tensor<1x1x167xf32>, %arg1: tensor<167x64xf32>) -> tensor<1x1x64xf32> {
//CHECK-NEXT: %0 = "vhlo.dot_general_v1"(%arg0, %arg1) <{

  func.return %0, %1 : !shape.shape, !shape.shape
}

// -----

// CHECK-LABEL: get_real_shape
func.func @get_real_shape(%arg0: tensor<1x2xf32>) -> tensor<2xindex> {
  %0 = "tfr.cast"(%arg0) : (tensor<1x2xf32>) -> !tfr.tensor
  %1 = tfr.get_shape %0 -> !shape.shape
  %2 = shape.to_extent_tensor %1 : !shape.shape -> tensor<2xindex>
  func.return %2 : tensor<2xindex>
}

// -----

// Tests for ops with canonicalization patterns.

// CHECK-LABEL: get_real_shape
func.func @get_real_shape(%arg0: tensor<1x2xf32>) -> tensor<2xindex> {
  %0 = "tfr.cast"(%arg0) : (tensor<1x2xf32>) -> !tfr.tensor
  %1 = tfr.get_shape %0 -> !shape.shape
  %2 = shape.to_extent_tensor %1 : !shape.shape -> tensor<2xindex>
  func.return %2 : tensor<2xindex>

)mlir";

constexpr absl::string_view kModuleHybridQuantized = R"mlir(
  module {
    func.func @main(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03:0>> {tf_saved_model.index_path = ["input_tensor"]}) -> (tensor<1x3xf32>) {
      %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x2xf32>, tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03:0>>) -> tensor<1x3xf32>
      return %0 : tensor<1x3xf32>
    }

  func.return %0 : tensor<2x2xbf16>

  // CHECK-LABEL: DivBf16LHS
  // CHECK: tf.Div
}

func.func @DontRemoveTrivialAdd(%arg0: tensor<1x2xf32>) -> tensor<2x2xf32> {
  %cst = arith.constant dense<0.0> : tensor<2x2xf32>
  %0 = "tf.AddV2"(%arg0, %cst) : (tensor<1x2xf32>, tensor<2x2xf32>) -> tensor<2x2xf32>
  func.return %0 : tensor<2x2xf32>

  // CHECK-LABEL: DontRemoveTrivialAdd