ldr.SetAttrSpecial(enoptrdata.Sym(), false)

	types := ldr.CreateSymForUpdate("runtime.types", 0)
	types.SetType(sym.STYPE)
	types.SetSize(8)
	ldr.SetAttrSpecial(types.Sym(), false)

	etypes := ldr.CreateSymForUpdate("runtime.etypes", 0)
	etypes.SetType(sym.SFUNCTAB)
	ldr.SetAttrSpecial(etypes.Sym(), false)

	if ctxt.HeadType == objabi.Haix {

  }

  // Infers shape of tf.While/tf.WhileRegion. If `shape_invariant` attribute is
  // set, operand types are set as result types if associated body result types
  // match the operand type (does not change per loop iteration). If operand and
  // body result types are not the same, only handle types are propagated to
  // result types. This is necessary to not incorrectly change result shapes

		terminatingErr          error
		prepareWorker           func(t *testing.T, w *podWorkers, records map[types.UID][]syncPodRecord)
		wantWorker              func(t *testing.T, w *podWorkers, records map[types.UID][]syncPodRecord)
		wantWorkerAfterRetry    func(t *testing.T, w *podWorkers, records map[types.UID][]syncPodRecord)
		wantErr                 bool
		expectMetrics           map[string]string

  return std::equal(i1, e1, i2);
}

// This function removes explicit broadcasting on type1 and returns whether if
// the reduced `type1` dimensions are the same as the ending dimensions
// of `type2`.
bool IsReducedTailOfShape(Type type1, Type type2) {
  auto tail_type = mlir::dyn_cast<ShapedType>(type1);
  auto full_type = mlir::dyn_cast<ShapedType>(type2);
  if (!tail_type || !full_type || !tail_type.hasRank() || !full_type.hasRank())

      ('use_constant_with_int32_input', dtypes.int32, False, True),
      ('use_variable_with_int32_input', dtypes.int32, True, True),
      ('use_constant_with_int64_input', dtypes.int64, False, True),
      ('use_variable_with_int64_input', dtypes.int64, True, True),
      ('small_gather_use_constant', dtypes.int32, False, False),
      ('small_gather_use_variable', dtypes.int32, True, False),
  )
  @test_util.run_v2_only

Attribute ConstFoldBinaryOp(
    Type result_type, ArrayRef<Attribute> operands,
    llvm::function_ref<APFloat(APFloat, APFloat)> float_calculate,
    llvm::function_ref<APInt(APInt, APInt)> int_calculate) {
  // Note: All types are wrapped in tensor types in TFlite. E.g., f32 is
  // represented as tensor<f32>. So we are only handling tensor types here.
  auto type = result_type.dyn_cast<ShapedType>();
  if (!type) return {};

	if expect, actual := []types.UID(nil), kubelet.podWorkers.(*fakePodWorkers).triggeredDeletion; !reflect.DeepEqual(expect, actual) {
		t.Fatalf("expected %v kills, got %v", expect, actual)
	}

	ready = true
	kubelet.HandlePodRemoves(pods)
	time.Sleep(2 * time.Second)

	// Sources are ready. Remove unwanted pods.

						Succeeded: func() []types.UID {
							uids := make([]types.UID, 499)
							for i := range uids {
								uids[i] = types.UID(strconv.Itoa(i))
							}
							return uids
						}(),
						Failed: []types.UID{"b"},
					},
					Failed: 1,
				},
				{
					UncountedTerminatedPods: &batch.UncountedTerminatedPods{
						Failed: []types.UID{"b"},
					},
					Succeeded: 499,

                    Attribute value) {
  // Handle the case where the type and value are already tensors.
  if (mlir::isa<TensorType>(type) && mlir::isa<ElementsAttr>(value)) {
    result.addTypes(type);
    result.addAttribute("value", value);
    return;
  }

  // Otherwise, default to the attribute builder.
  ConstOp::build(builder, result, value);

        VLOG(1) << "[potentially conservative] Op type `" << node.type_string()
                << "` is stateful but effects not modelled";
      } else {
        // See if any resource type is used.
        bool resource = false;
        std::function<bool(mlir::Type)> record_resource;
        record_resource = [&](mlir::Type type) {
          type.walk([&](mlir::Type t) {