## Bug Fixes and Other Changes

*   Use threadsafe version of `lgamma` function.
*   Fix `tf.sqrt` handling of negative arguments.
*   Fixed bug causing incorrect number of threads to be used for multi-threaded
    benchmarks.
*   Performance optimizations for `batch_matmul` on multi-core CPUs.
*   Improve trace, `matrix_set_diag`, `matrix_diag_part` and their gradients to
    work for rectangular matrices.

            }
        }
        if (nThreads < 1) {
            nThreads = 1;
        }
        final ExecutorService executorService = new ThreadPoolExecutor(nThreads, nThreads, 0L, TimeUnit.MILLISECONDS,
                new LinkedBlockingQueue<Runnable>(nThreads), new ThreadPoolExecutor.CallerRunsPolicy());
        try {
            while (!isInterrupted() && !stopIfLeisure()) {

    static final int O_TRUNC = 0x0040;

    // share access
    /**
     * When specified as the {@code shareAccess} constructor parameter,
     * other SMB clients (including other threads making calls into jCIFS)
     * will not be permitted to access the target file and will receive "The
     * file is being accessed by another process" message.
     */
    public static final int FILE_NO_SHARE = 0x00;

     *
     * @return the {@link CrawlerContext} for the current thread, or {@code null} if no context is set.
     */
    public static CrawlerContext getCrawlerContext() {
        return CRAWLER_CONTEXT_THREAD_LOCAL.get();
    }

    /**
     * Sets the current {@link CrawlerContext} for the current thread.
     * If the provided {@code crawlerContext} is {@code null}, the context is removed from the thread-local storage.

  @GuardedBy("queue")
  private WorkerRunningState workerRunningState = IDLE;

  /**
   * This counter prevents an ABA issue where a thread may successfully schedule the worker, the
   * worker runs and exhausts the queue, another thread enqueues a task and fails to schedule the
   * worker, and then the first thread's call to delegate.execute() returns. Without this counter,

  @GuardedBy("queue")
  private WorkerRunningState workerRunningState = IDLE;

  /**
   * This counter prevents an ABA issue where a thread may successfully schedule the worker, the
   * worker runs and exhausts the queue, another thread enqueues a task and fails to schedule the
   * worker, and then the first thread's call to delegate.execute() returns. Without this counter,

- "unfinished_work_microseconds" is added to the workqueue metrics; it can be used to detect stuck worker threads. (kube-controller-manager runs many workqueues.) ([#70884](https://github.com/kubernetes/kubernetes/pull/70884), [@lavalamp](https://github.com/lavalamp))

     *       observing a null thread does not mean the object is safe to reuse.
     *   <li>If this field's value is some other thread object, we know that it's not our thread.
     *   <li>If this field's value == null because it originally belonged to another thread and that
     *       thread cleared it, we still know that it's not associated with our thread

 *
 *  * For asynchronous calls made with [Call.enqueue], an [IOException] is propagated to the caller
 *    indicating that the call was canceled. The interceptor's exception is delivered to the current
 *    thread's [uncaught exception handler][Thread.UncaughtExceptionHandler]. By default this
 *    crashes the application on Android and prints a stacktrace on the JVM. (Crash reporting
 *    libraries may customize this behavior.)
 *

     *       observing a null thread does not mean the object is safe to reuse.
     *   <li>If this field's value is some other thread object, we know that it's not our thread.
     *   <li>If this field's value == null because it originally belonged to another thread and that
     *       thread cleared it, we still know that it's not associated with our thread