func.func @rng_bit_generator(%arg0: tensor<2xui64>) -> (tensor<2xui64>, tensor<10x12xui32>) {
  %output_state, %output = "vhlo.rng_bit_generator_v1"(%arg0) <{rng_algorithm = #vhlo<rng_algorithm_v1 DEFAULT>}> : (tensor<2xui64>) -> (tensor<2xui64>, tensor<10x12xui32>)
  func.return %output_state, %output : tensor<2xui64>, tensor<10x12xui32>
}

// CHECK:func.func private @rng_bit_generator(%arg0: tensor<2xui64>) -> (tensor<2xui64>, tensor<10x12xui32>) {

  %0 = mhlo.composite "aten.avg_pool2d.default" %arg0 {composite_attributes = {ceil_mode = false, count_include_pad = true, divisor_override = "py_None", kernel_size = dense<3> : tensor<2xi64>, padding = dense<0> : tensor<2xi64>, stride = dense<1> : tensor<2xi64>}, decomposition = @XlaCallModule_aten.avg_pool2d.default.impl_0} : (tensor<1x3x6x6xf32>) -> tensor<1x3x4x4xf32>
  %1 = "tf.Identity"(%0) {device = ""} : (tensor<1x3x4x4xf32>) -> tensor<*xf32>

func.func @stridedSliceGrad(%dy: tensor<4x8xf32>, %begin: tensor<2xi64>, %end: tensor<2xi64>, %strides: tensor<2xi64>, %shape: tensor<2xi64>) -> tensor<?x?xf32> {
  // CHECK: tf.StridedSliceGrad
  %0 = "tf.StridedSliceGrad"(%shape, %begin, %end, %strides, %dy) : (tensor<2xi64>, tensor<2xi64>, tensor<2xi64>, tensor<2xi64>, tensor<4x8xf32>) -> tensor<?x?xf32>
  func.return %0 : tensor<?x?xf32>
}

// -----

  %32 = "tf.ConcatV2"(%29#0, %31, %14) {device = ""} : (tensor<2xi64>, tensor<1xi64>, tensor<i32>) -> tensor<3xi64>
  %33 = "tf.GatherV2"(%24#2, %29#1, %14) {batch_dims = 0 : i64, device = ""} : (tensor<?xi64>, tensor<?xi64>, tensor<i32>) -> tensor<?xi64>
  %34 = "tf.ConcatV2"(%33, %22, %14) {device = ""} : (tensor<?xi64>, tensor<1xi64>, tensor<i32>) -> tensor<?xi64>

  // CHECK-DAG: %[[MATMUL_LHS_SHAPE:.*]] = "tf.Const"() <{value = dense<[4, 5]> : tensor<2xi64>}> : () -> tensor<2xi64>
  // CHECK-DAG: %[[MATMUL_RHS_SHAPE:.*]] = "tf.Const"() <{value = dense<[5, 6]> : tensor<2xi64>}> : () -> tensor<2xi64>

  // CHECK: %[[LHS_1:.*]] = "tf.Reshape"(%arg0, %[[MATMUL_LHS_SHAPE]]) : (tensor<1x4x5xf32>, tensor<2xi64>) -> tensor<4x5xf32>

  // CHECK-NEXT: %12 = "mhlo.pad"(%10, %11) <{edge_padding_high = dense<[3, 4]> : tensor<2xi64>, edge_padding_low = dense<0> : tensor<2xi64>, interior_padding = dense<0> : tensor<2xi64>}> : (tensor<3x4xi32>, tensor<i32>) -> tensor<6x8xi32>
  // CHECK-NEXT: %13 = "mhlo.slice"(%arg1) <{limit_indices = dense<1> : tensor<1xi64>, start_indices = dense<0> : tensor<1xi64>, strides = dense<1> : tensor<1xi64>}> : (tensor<2xi32>) -> tensor<1xi32>

// CHECK-DAG:       %[[VAL_2:.*]] = "tf.Const"() <{value = dense<[1, 0]> : tensor<2xi64>}> : () -> tensor<2xi64>
// CHECK-DAG:       %[[VAL_3:.*]] = "tf.Const"() <{value = dense<[1, 0]> : tensor<2xi64>}> : () -> tensor<2xi64>
// CHECK:           %[[VAL_4:.*]] = "tf.Transpose"(%[[VAL_0]], %[[VAL_3]]) : (tensor<2x3xf32>, tensor<2xi64>) -> tensor<3x2xf32>

// CHECK: %[[arg2_i64:.*]] = "tf.Cast"(%arg2) <{Truncate = false}> : (tensor<2xi32>) -> tensor<2xi64>
// CHECK: %[[slice:.*]] = "tf.Slice"(%arg0, %[[tensor_scatter_update]], %[[arg2_i64]]) : (tensor<?x2xf32>, tensor<2xi64>, tensor<2xi64>) -> tensor<*xf32>
// CHECK: %[[reshape:.*]] = "tf.Reshape"(%[[slice]], %[[cst_1]]) : (tensor<*xf32>, tensor<1xi64>) -> tensor<*xf32>

  // CHECK: mhlo.reshape {{.*}} : (tensor<2x1xi64>) -> tensor<2xi64>
  // CHECK: "mhlo.slice"({{.*}}) <{limit_indices = dense<1> : tensor<1xi64>, start_indices = dense<0> : tensor<1xi64>, strides = dense<1> : tensor<1xi64>}> : (tensor<2xi64>) -> tensor<1xi64>

// CHECK-NEXT:      %16 = "tfl.pseudo_const"() <{value = dense<[0, 1]> : tensor<2xi64>}> : () -> tensor<2xi64>
// CHECK-NEXT:      %17 = "tfl.gather"(%14, %16) <{axis = 0 : i32, batch_dims = 0 : i32}> : (tensor<3xi32>, tensor<2xi64>) -> tensor<2xi32>
// CHECK-NEXT:      %18 = "tfl.pseudo_const"() <{value = dense<1> : tensor<1xi64>}> : () -> tensor<1xi64>