func (e errorString) Error() string {
	return "runtime error: " + string(e)
}

var overflowError = error(errorString("integer overflow"))

  }

  // Call AddDevices to register the XLA devices.
  //
  // It may be worth refactoring out XlaOpRegistry::RegisterCompilationDevice to
  // make this more direct, but probably not worth it solely for this test.
  std::vector<std::unique_ptr<Device>> devices;
  TF_RETURN_IF_ERROR(DeviceFactory::AddDevices(session_options, "", &devices));

  GraphOptimizationPassOptions opt_options;
  opt_options.graph = graph;

//
// Input:
//     Tensorflow Dialect MLIR with tf_device.cluster ops and virtual devices.
//     xla_device_type - The device type that is being targeted.
// Output:
//     Tensorflow Dialect MLIR with Runtime specific ops. All tf_device.cluster
//     ops are removed. Physical devices are assigned to ops instead of virtual
//     devices.
tensorflow::Status RunLowerClusterToRuntimeOpsPassPipeline(

// configuring the persistor used in the DeviceCompiler. Please note that
// non-XLA devices aren't supported yet. This is because:
// 1. PjRtClient doesn't support data transfer for non-XLA devices yet
// 2. Fetching the PjRtClient for non-XLA devices is also not supported yet
Status GetOrCreatePjRtDeviceCompilerAndProfiler(
    const OpKernelContext& ctx, const XlaPlatformInfo& platform_info,

  // CHECK-SAME: (%[[ARG0]] as %[[RI_0:[a-z0-9]*]]: tensor<*x!tf_type.resource<tensor<4xf32>>>)
  tf_device.replicate(%arg0 as %arg1: tensor<*x!tf_type.resource<tensor<4xf32>>>) {
    _mirrored_variable_indices = [0],
    devices = {TPU_REPLICATED_CORE_0 = ["/job:worker/replica:0/task:0/device:TPU:0", "/job:worker/replica:0/task:0/device:TPU:1"]},
    n = 2 : i32} {
     // CHECK:      %[[RESOURCE_OUT:.*]] = "tf_device.launch"()

    auto device_factory = DeviceFactory::GetFactory(device_type);
    SessionOptions options;
    std::vector<std::unique_ptr<Device>> devices;
    Status s = device_factory->CreateDevices(
        options, "/job:worker/replica:0/task:0", &devices);
    device_ = std::move(devices[0]);

    tensorflow::AllocatorAttributes host_alloc_attr;
    host_alloc_attr.set_on_host(true);

    std::function<WalkResult(Operation*, tf_device::ClusterOp,
                             std::optional<std::string>)>
        callback) {
  mlir::TF::RuntimeDevices devices;
  if (failed(tensorflow::GetDevicesFromOp(module, &devices))) return failure();
  const CallGraph call_graph(module);
  // symbol_table caches callees in the CallGraph.
  SymbolTableCollection symbol_table;

// RUN: tf-tfrt-opt -tfrt-lower-cluster-to-runtime-ops-tpu -split-input-file -verify-diagnostics %s | FileCheck %s

module attributes {tf.versions = {producer = 888 : i32}, tf.devices = ["/job:worker/replica:0/task:0/device:CPU:0", "/job:worker/replica:0/task:0/device:TPU_SYSTEM:0", "/job:worker/replica:0/task:0/device:TPU:0"]} {

  // CHECK-LABEL: @converts_cluster
  func.func @converts_cluster() {
    // CHECK: %0:2 = "tf_device.launch"() <{{.*}}> ({

	return 1<<63 - 1 + int64(unsafe.Sizeof(*new(T)))
}

func main() {
	shift[[62]byte]()
	shift[[63]byte]()
	shift[[64]byte]()
	shift[[100]byte]()
	shift[[1e6]byte]()

	add[[1]byte]()
	shouldPanic("divide by zero", func() { div[[0]byte]() })
}

func shouldPanic(str string, f func()) {
	defer func() {
		err := recover()
		if err == nil {
			panic("did not panic")
		}
		s := err.(error).Error()

const DefaultPhysPageSize = goarch.DefaultPhysPageSize

// PCQuantum is the minimal unit for a program counter (1 on x86, 4 on most other systems).
// The various PC tables record PC deltas pre-divided by PCQuantum.
const PCQuantum = goarch.PCQuantum

// Int64Align is the required alignment for a 64-bit integer (4 on 32-bit systems, 8 on 64-bit).
const Int64Align = goarch.PtrSize