// Test valid tf.Select
// CHECK-LABEL: func @testSelect
func.func @testSelect(%arg0: tensor<3xi1>, %arg1: tensor<3x2xf16>, %arg2: tensor<3x2xf16>) -> tensor<3x2xf16> {
  %0 = "tf.Select"(%arg0, %arg1, %arg2) : (tensor<3xi1>, tensor<3x2xf16>, tensor<3x2xf16>) -> tensor<3x2xf16>
  func.return %0: tensor<3x2xf16>
}

// -----

}

func.func @floor_mod_i16(%arg0: tensor<5xi16>, %arg1: tensor<5xi16>) -> tensor<5xi16> {
  %0 = "tf.FloorMod"(%arg0, %arg1) : (tensor<5xi16>, tensor<5xi16>) -> tensor<5xi16>
  func.return %0 : tensor<5xi16>
  // CHECK-LABEL: floor_mod_i16
  // CHECK: "tfl.floor_mod"(%arg0, %arg1) : (tensor<5xi16>, tensor<5xi16>) -> tensor<5xi16>
}

func.func @exp(%arg0: tensor<5xf32>) -> tensor<5xf32> {

  module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    func.func @main() -> () {
      %cst0 = "tf.Const"(){ value = dense<0> : tensor<3x5xi1>} : () -> tensor<3x5xi1>
      %0 = "tf.Where"(%cst0) : (tensor<3x5xi1>) -> tensor<?x2xi64>
      func.return
    }
  })";

  CellReader<int64_t> dynamism_counter(kDynamismOpCounterStreamzName);
  CellReader<int64_t> dynamism_function_counter(

  func.return %0: tensor<3x2xf16>
}

func.func @sparse_qu8() -> tensor<3x2x!quant.uniform<u8<1:255>:f32, 1.0>> {
  // CHECK-LABEL: @sparse_qu8

  func.return %0 : tensor<3x4xi1>
}

// CHECK-LABEL: @is_nan
func.func @is_nan(%arg0: tensor<3x4xf32>) -> tensor<3x4xi1> {
  // CHECK: %[[RESULT:.*]] = "tf.NotEqual"(%arg0, %arg0) <{incompatible_shape_error = true}> : (tensor<3x4xf32>, tensor<3x4xf32>) -> tensor<3x4xi1>
  %0 = "tf.IsNan"(%arg0) : (tensor<3x4xf32>) -> tensor<3x4xi1>
  // CHECK: return %[[RESULT]]
  func.return %0 : tensor<3x4xi1>
}

func.func @broadcast_to_i16_low_dim(%input: tensor<3xi16>, %shape: tensor<2xi32>) -> tensor<3x3xi16> {
  %0 = "tf.BroadcastTo"(%input, %shape) : (tensor<3xi16>, tensor<2xi32>) -> tensor<3x3xi16>
  func.return %0: tensor<3x3xi16>

// CHECK-LABEL: broadcast_to_i16_low_dim
// CHECK:    %0 = "tf.BroadcastTo"(%arg0, %arg1) : (tensor<3xi16>, tensor<2xi32>) -> tensor<3x3xi16>
// CHECK:    return %0 : tensor<3x3xi16>
}

  func.return %0: tensor<3x2xf16>
}

func.func @sparse_qu8() -> tensor<3x2x!quant.uniform<u8<1:255>:f32, 1.0>> {
  // CHECK-LABEL: @sparse_qu8

  module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
  func.func @main() -> () {
    %cst0 = "tf.Const"(){ value = dense<0> : tensor<3x5xi1>} : () -> tensor<3x5xi1>
    %0 = "tf.Where"(%cst0) : (tensor<3x5xi1>) -> tensor<?x2xi64>
    func.return
  }
})";

absl::StatusOr<XlaCompiler::CompilationResult> CompileMlirModule(
    const char* mlir_module_str,

func.func @broadcast_to_i16_low_dim(%arg0: tensor<3xi16>, %arg1: tensor<2xi32>) -> tensor<3x3xi16> {
  %0 = "tfl.broadcast_to"(%arg0, %arg1) : (tensor<3xi16>, tensor<2xi32>) -> tensor<3x3xi16>
  return %0 : tensor<3x3xi16>
  // CHECK:  %cst = arith.constant dense<1> : tensor<3x3xi16>
  // CHECK:  %0 = tfl.mul(%arg0, %cst) <{fused_activation_function = "NONE"}> : (tensor<3xi16>, tensor<3x3xi16>) -> tensor<3x3xi16>
  // CHECK:  return %0 : tensor<3x3xi16>

// CHECK-LABEL: test_gather
func.func @test_gather(%arg0 : tensor<1x2x20xf32>, %arg1 : tensor<3x5xi32>) -> (tensor<1x3x5x20xf32>){
    // expected-remark@+1 {{Cost: 1}}
    %0 = "tf.Const"() { value = dense<[1]> : tensor<1xi32> } : () -> tensor<1xi32>
    // expected-remark@+1 {{Cost: 300}}
    %1 = "tf.GatherV2"(%arg0, %arg1, %0) : (tensor<1x2x20xf32>, tensor<3x5xi32>, tensor<1xi32>) -> tensor<1x3x5x20xf32>
    // expected-remark@+1 {{Cost: 40}}