// CHECK:           [[VAL_40:%.*]] = arith.constant dense<0> : tensor<1xi32>
// CHECK:           [[VAL_41:%.*]] = "tf.ReverseV2"([[VAL_0]], [[VAL_40]]) : (tensor<8x8x8xf32>, tensor<1xi32>) -> tensor<8x8x8xf32>
// CHECK:           [[VAL_6:%.*]] = arith.constant dense<[1, 0]> : tensor<2xi32>
// CHECK:           [[VAL_7:%.*]] = "tf.Transpose"([[VAL_3]], [[VAL_6]]) : (tensor<8x40xf32>, tensor<2xi32>) -> tensor<40x8xf32>

      tf_device.return %2 : tensor<1x8x2xf32>
    // CHECK: () -> tensor<1x8x2xf32>
    }) {device = "/device:CPU:0"} : () -> tensor<*xf32>
    // CHECK: "tf.Cast"(%{{.*}}) <{Truncate = false}> : (tensor<1x8x2xf32>) -> tensor<*xf32>
    // CHECK: (tensor<i32>, tensor<1x8x2xf32>) -> (tensor<1x8x1xf32>, tensor<1x8x1xf32>)

  } : (tensor<9x10x!tf_type.qint8>, tensor<f32>, tensor<i32>) -> tensor<9x10xf32>
  %0 = "tf.MatMul"(%input, %filter_new) {
  } : (tensor<8x9xf32>, tensor<9x10xf32>) -> tensor<8x10xf32>
  return %0 : tensor<8x10xf32>
})mlir";
  constexpr absl::string_view kProgram = R"mlir(
func.func @main(%input: tensor<8x9xf32>, %filter: tensor<9x10xi8>) -> tensor<8x10xf32> {
  %filter_scale = "tf.Const"() { value = dense<0.0235> : tensor<f32> } : ()

}

func.func @scatterNdVectorIndices(%arg0: tensor<5x1xi32>, %arg1: tensor<5x3x2xf32>) -> tensor<10x3x2xf32> {
  %cst = "tf.Const"() { value = dense<[10, 3, 2]> : tensor<3xi32> } : () -> tensor<3xi32>
  %1 = "tf.ScatterNd"(%arg0, %arg1, %cst) : (tensor<5x1xi32>, tensor<5x3x2xf32>, tensor<3xi32>) -> tensor<10x3x2xf32>
  func.return %1 : tensor<10x3x2xf32>

// CHECK-LABEL:scatterNdVectorIndices

func.func @testDilate(%arg0: tensor<3x4x5xf32>) -> tensor<5x7x9xf32> {
  // CHECK: "tfl.dilate"(%arg0, %cst, %cst_0)
  %cst = arith.constant dense<1> : tensor<3xi32>
  %cst_0 = arith.constant dense<-1.0> : tensor<f32>
  %0 = "tfl.dilate"(%arg0, %cst, %cst_0) : (tensor<3x4x5xf32>, tensor<3xi32>, tensor<f32>) -> tensor<5x7x9xf32>
  func.return %0 : tensor<5x7x9xf32>
  // CHECK: return %0 : tensor<5x7x9xf32>

// CHECK:  return %[[RESULT]]
}

func.func @ConvertSqueezeToReshapeWithDynamicDimension2(%arg0: tensor<?x1x8x3xf32>) -> tensor<1x8x3xf32> {
  %0 = "tfl.squeeze"(%arg0) {squeeze_dims = [0]}: (tensor<?x1x8x3xf32>) -> tensor<1x8x3xf32>
  func.return %0: tensor<1x8x3xf32>

// CHECK-LABEL: ConvertSqueezeToReshapeWithDynamicDimension2

  %0 = "tf.InplaceUpdate"(%arg0, %arg2, %arg1) : (tensor<8x8x4xf32>, tensor<3xi32>, tensor<3x8x4xf32>) -> tensor<8x8x4xf32>

  // CHECK:  return [[UPDATE3]] : tensor<8x8x4xf32>
  func.return %0 : tensor<8x8x4xf32>
}

// -----

// CHECK-LABEL: xla_dynamic_update_slice

// CHECK:           return %[[VAL_7]] : tensor<5x7x9xi32>
// CHECK:         }
func.func @convert_iota_3d() -> tensor<5x7x9xi32> {
  %0 = "mhlo.iota"() <{ iota_dimension = 1 : i64 }> : () -> tensor<5x7x9xi32>
  func.return %0 : tensor<5x7x9xi32>
}

// CHECK-LABEL:   func @convert_avgpool_valid(

  %arg0_shape = "tf.Const"() {value = dense<[1, 8, 4]> : tensor<3xi32>} : () -> tensor<3xi32>
  %arg0_reshaped = "tf.Reshape"(%arg0, %arg0_shape) : (tensor<8x4xf32>, tensor<3xi32>) -> tensor<1x8x4xf32>
  %zeroi2 = "tf.Const"() {value = dense<0> : tensor<2xi32>} : () -> tensor<2xi32>
  %axis = "tf.Const"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>