{
    mutex_lock l(mu_);
    if (should_stop_) {
      return Status(absl::StatusCode::kFailedPrecondition,
                    "The coordinator has been stopped.");
    }
  }
  mutex_lock l(runners_lock_);
  runners_.push_back(std::move(runner));
  return absl::OkStatus();
}

bool Coordinator::AllRunnersStopped() {
  mutex_lock l(runners_lock_);
  for (const auto& runner : runners_) {

Status QueueRunner::Join() {
  thread_pool_.reset();
  mutex_lock l(mu_);
  return status_;
}

void QueueRunner::UpdateStatus(const Status& status) {
  {
    mutex_lock l(mu_);
    if (!status_.ok() || status.ok() || IsQueueClosed(status)) {
      return;
    }
    status_ = status;
  }
  if (coord_) {
    coord_->ReportStatus(status);
  }
  mutex_lock l(cb_mu_);
  for (auto& cb : callbacks_) {

}

void DeviceCompilationProfiler::IncrementOngoingAsyncCompilations() {
  mutex_lock lock(mu_);
  num_ongoing_compilations_++;
}

void DeviceCompilationProfiler::DecrementOngoingAsyncCompilations() {
  mutex_lock lock(mu_);
  num_ongoing_compilations_--;
}

int64_t DeviceCompilationProfiler::GetNumOngoingAsyncCompilations() const {
  mutex_lock lock(mu_);
  return num_ongoing_compilations_;
}

  // protect the contents of the cache entry.
  Entry* entry;
  {
    mutex_lock lock(compile_cache_mu_);
    // Find cache entry.
    auto it = cache_.find(key);
    if (it == cache_.cend()) {
      return std::nullopt;
    }

    entry = it->second.get();
  }

  mutex_lock lock(entry->mu);
  Value value = {/*compile_state=*/entry->compile_state,

}

void XlaTensor::ResetDefinitionEvent(std::shared_ptr<se::Event> event,
                                     se::Stream* stream) {
  mutex_lock lock(mu_);
  definition_event_ = std::move(event);
  streams_defined_on_ = {stream};
}

Status XlaTensor::RefreshStatusOfStreams() {
  mutex_lock lock(mu_);
  Status status;
  for (se::Stream* stream : streams_defined_on_) {
    status.Update(stream->RefreshStatus());
  }

#include "tensorflow/core/public/session.h"

namespace mlir {
namespace tf_saved_model {

using llvm::SmallSet;
using ::tensorflow::Device;
using ::tensorflow::DeviceMgr;
using ::tensorflow::mutex_lock;
using ::tensorflow::ResourceHandle;
using ::tensorflow::Session;
using ::tensorflow::Status;
using ::tensorflow::StatusOr;
using ::tensorflow::Tensor;
using ::tensorflow::Var;

namespace {

                                *(*ptr)->tensor() = value;
                                (*ptr)->is_initialized = true;
                                return absl::OkStatus();
                              }));
  mutex_lock ml(*variable->mu());
  OP_REQUIRES(
      context,
      !variable->is_initialized || variable->tensor()->dtype() == dtype_,
      errors::InvalidArgument(
          "Trying to assign variable with wrong dtype. Expected ",

                          std::vector<Tensor>* outputs) const {
  return Run(RunOptions(), inputs, fetch_outputs, run_outputs, outputs,
             nullptr);
}

Status ClientSession::Impl::MaybeExtendGraph() const {
  mutex_lock l(mu_);
  int num_nodes = graph_->num_node_ids();
  if (num_nodes > last_num_graph_nodes_) {
    GraphDef graph_def;
    graph_->ToGraphDefSubRange(&graph_def, last_num_graph_nodes_);
    last_num_graph_nodes_ = num_nodes;

// while allowing other goroutines to run.

const (
	mutex_unlocked = 0
	mutex_locked   = 1

	active_spin     = 4
	active_spin_cnt = 30
)

func mutexContended(l *mutex) bool {
	return false
}

func lock(l *mutex) {
	lockWithRank(l, getLockRank(l))
}

func lock2(l *mutex) {
	if l.key == mutex_locked {
		// wasm is single-threaded so we should never
		// observe this.
		throw("self deadlock")

//	futexwakeup(addr *uint32, cnt uint32)
//		If any procs are sleeping on addr, wake up at most cnt.

const (
	mutex_unlocked = 0
	mutex_locked   = 1
	mutex_sleeping = 2

	active_spin     = 4
	active_spin_cnt = 30
	passive_spin    = 1
)

// Possible lock states are mutex_unlocked, mutex_locked and mutex_sleeping.
// mutex_sleeping means that there is presumably at least one sleeping thread.