// RUN: tf-quant-opt %s -split-input-file -quant-preprocess-op -quant-prepare-quantize-drq | FileCheck %s

module {
  func.func @matmul(%arg0: tensor<1x2x2x3xf32>) -> (tensor<*xf32>) {
    %cst_0 = "tf.Const"() {value = dense<0.000000e+00> : tensor<2x1024xf32>} : () -> tensor<2x1024xf32>

  return %0 : tensor<1x3xf32>
}

// CHECK-LABEL: @_func
func.func private @_func(%arg0: tensor<1x3xf32>) -> (tensor<1x3xf32>) {
  return %arg0 : tensor<1x3xf32>
}
}


// -----
// SPMD: one input and no return
//
// CHECK-LABEL: func.func @serving_default(%arg0: tensor<1x3xf32>) {
// CHECK-NEXT:  "tf.IfrtCall"(%arg0)
// CHECK-SAME:       {program_id = [[PROGRAM_ID:.*]] : i64, variable_arg_indices = []}

    // CHECK-NEXT:   %[[C_OUT:.*]] = "tf.C"(%arg0, %[[CLUSTER_OUT]]#1)
    // CHECK-NOT:    _xla_outside_compilation
    // CHECK-NEXT:   "tf.D"(%[[C_OUT]], %arg0, %[[CLUSTER_OUT]]#0)
    // CHECK-NOT:    _xla_outside_compilation
    // CHECK-NEXT:   tf_device.return
    "tf_device.cluster"() ({
      %a = "tf.A"() : () -> tensor<i32>
      %b = "tf.B"(%arg0) : (tensor<i32>) -> tensor<i32>

// CHECK-NOT: tf._noinline
func.func @inner_stateful_pcall_func(%arg0: tensor<i32>) -> tensor<i32> {
  %0 = "tf.StatefulPartitionedCall"(%arg0) {_xla_compile_device_type = "CPU", config = "", config_proto = "", device = "/device:CPU:0", executor_type = "", f = @func} : (tensor<i32>) -> (tensor<i32>)
  func.return %0 : tensor<i32>
}

func.func @func(%arg0: tensor<i32>) -> tensor<i32> {
  func.return %arg0 : tensor<i32>

// CHECK:           return %[[VAL_3]] : tensor<2xf32>
// CHECK:         }
func.func @floordiv_f32(%arg0: tensor<2xf32>) -> tensor<2xf32> {
  %0 = mhlo.divide %arg0, %arg0 : tensor<2xf32>
  %1 = mhlo.divide %arg0, %arg0 : tensor<2xf32>
  %2 = "mhlo.floor"(%1) : (tensor<2xf32>) -> tensor<2xf32>
  func.return %2 : tensor<2xf32>
}

// CHECK-LABEL:   func @floordiv_f16_broadcast(

// CHECK-LABEL: func private @batched_function
func.func private @batched_function(%arg0: tensor<1x3xf32>) -> tensor<1x3xf32> {
  %2 = "tf.Identity"(%arg0) : (tensor<1x3xf32>) -> tensor<1x3xf32>
  func.return %2 : tensor<1x3xf32>
}

// CHECK-LABEL: func @main
func.func @main(%arg0: tensor<1x3xf32>) -> tensor<*xf32> {
  // CHECK:  "tf.BatchFunction"
  // CHECK-SAME: allowed_batch_sizes = [6]

// CHECK-LABEL: nchw_conv
// CHECK-SAME: %[[ARG:.+]]: tensor<1x8x4x4xf32>
func.func @nchw_conv(%arg0: tensor<1x8x4x4xf32>) -> tensor<1x8x4x4xf32> {
  %0 = stablehlo.constant() {value = dense<7.000000e+00> : tensor<8x8x3x3xf32>} : () -> tensor<8x8x3x3xf32>

// CHECK-LABEL: func @test_supported_lowering_of_tf_if_region1
// CHECK-SAME: (%[[ARG0:.*]]: tensor<i1>, %[[ARG1:.*]]: tensor<4xf32>)
func.func @test_supported_lowering_of_tf_if_region1(%arg0: tensor<i1>, %arg1: tensor<4xf32>) -> (tensor<*xf32>, tensor<4xf32>) {
  %res:2 = "tf.IfRegion"(%arg0) ({
    %call = func.call @test_if_then1(%arg1) : (tensor<4xf32>) -> tensor<4xf32>

  // CHECK-LABEL: func @annotate_broadcast_values
  func.func @annotate_broadcast_values(%arg0: tensor<?xi32>) -> tensor<?xi32> {
    %0 = "tf._A"(%arg0) : (tensor<?xi32>) -> tensor<?xi32>
    %1 = "tf._B"(%arg0) : (tensor<?xi32>) -> tensor<?xi32>
    %5:2 = tf_device.replicate([%0, %arg0] as %ri_0: tensor<?xi32>) {n = 2 : i32} {
      %2 = "tf._F"(%arg0) : (tensor<?xi32>) -> tensor<?xi32>
      %3 = "tf.Identity"(%1) : (tensor<?xi32>) -> tensor<?xi32>

}

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

  // CHECK-LABEL: RemoveTrivialAdd
  // CHECK-NEXT: return %arg0 : tensor<2x2xf32>
}

func.func @RemoveTrivialAddBf16RHS(%arg0: tensor<2x2xbf16>) -> tensor<2x2xbf16> {