return nil
	}
	metrics.ObserveQueueLength(ctx, qs.qCfg.Name, fsName, queue.requestsWaiting.Length())
	return req
}

// shuffleShardLocked uses shuffle sharding to select a queue index
// using the given hashValue and the shuffle sharding parameters of the queueSet.
func (qs *queueSet) shuffleShardLocked(hashValue uint64, descr1, descr2 interface{}) int {
	var backHand [8]int

            }
            project(':lib3') {
                apply plugin: 'cpp-library'
            }
"""
        app.deck.writeToProject(file("lib1"))
        app.card.writeToProject(file("lib2"))
        app.shuffle.writeToProject(file("lib3"))

        expect:
        succeeds tasks(':lib1').debug.assemble

	MIX(a, b, c, d, e, 0x5A827999)

#define ROUND1x(a, b, c, d, e, index) \
	SHUFFLE(index); \
	FUNC1(a, b, c, d, e); \
	MIX(a, b, c, d, e, 0x5A827999)

#define ROUND2(a, b, c, d, e, index) \
	SHUFFLE(index); \
	FUNC2(a, b, c, d, e); \
	MIX(a, b, c, d, e, 0x6ED9EBA1)

#define ROUND3(a, b, c, d, e, index) \
	SHUFFLE(index); \
	FUNC3(a, b, c, d, e); \
	MIX(a, b, c, d, e, 0x8F1BBCDC)

mapreduce.fileoutputcommitter.algorithm.version=2 # Use the latest committer version
mapreduce.job.reduce.slowstart.completedmaps=0.99 # 99% map, then reduce
mapreduce.reduce.shuffle.input.buffer.percent=0.9 # Min % buffer in RAM
mapreduce.reduce.shuffle.merge.percent=0.9 # Minimum % merges in RAM
mapreduce.reduce.speculative=false # Disable speculation for reducing
mapreduce.task.io.sort.factor=999 # Threshold before writing to disk

    ]

    # Extend the representative dataset with np.ndarrays.
    repr_ds.extend([
        {'tensor_key': np.random.uniform(low=-1.0, high=1.0, size=(3, 3))}
        for _ in range(4)
    ])

    random.shuffle(repr_ds)

    with self.session() as sess:
      new_repr_ds = repr_dataset.replace_tensors_by_numpy_ndarrays(
          repr_ds, sess
      )

      # The resulting dataset should not contain any tf.Tensors.

            spec.resolveStrategy = Closure.DELEGATE_FIRST
            spec()
            children.add(childBuilder)
        }

        ProjectDescriptor build() {
            children.shuffle()
            def childDescriptors = children*.build() as Set<TestProjectDescriptor>
            def thisProject = new TestProjectDescriptor()
            thisProject.name = name

      indexes.add(index);
      expectedBuilder.put(index, expectedLargeDatasetPercentile(index));
    }
    Random random = new Random(770683168895677741L);
    Collections.shuffle(indexes, random);
    assertThat(percentiles().indexes(Ints.toArray(indexes)).compute(PSEUDORANDOM_DATASET))
        .comparingValuesUsing(QUANTILE_CORRESPONDENCE)
        .containsExactlyEntriesIn(expectedBuilder.buildOrThrow());

              return null;
            }
          });
    }
    assertEquals(allTasks.size() + 1, barrier.getParties());
    for (int i = 0; i < 1000; i++) {
      Collections.shuffle(allTasks);
      final AbstractFuture<String> future = new AbstractFuture<String>() {};
      currentFuture.set(future);
      for (Callable<?> task : allTasks) {

  // +optional
  optional int32 queues = 1;

  // `handSize` is a small positive number that configures the
  // shuffle sharding of requests into queues.  When enqueuing a request
  // at this priority level the request's flow identifier (a string
  // pair) is hashed and the hash value is used to shuffle the list
  // of queues and deal a hand of the size specified here.  The

  // +optional
  optional int32 queues = 1;

  // `handSize` is a small positive number that configures the
  // shuffle sharding of requests into queues.  When enqueuing a request
  // at this priority level the request's flow identifier (a string
  // pair) is hashed and the hash value is used to shuffle the list
  // of queues and deal a hand of the size specified here.  The