* The cache has a builtin test, enabled with `UNSAFE_PILOT_ENABLE_RUNTIME_ASSERTIONS=true`, that runs in CI. This will panic if any key is written to with a different value.

#### Partial Computations

Along with caching, partial computations are a critical performance optimization to ensure that we do not need to build (or send) every resource to every proxy on every change. This is discussed more in the Config Serving section.

Those worker processes would be the ones running your application, they would perform the main computations to receive a **request** and return a **response**, and they would load anything you put in variables in RAM.

<img src="/img/deployment/concepts/process-ram.svg">

    computationLatch.countDown();
    // let some computations complete
    while (computedCount.get() < nThreads) {
      Thread.yield();
    }
    cache.invalidateAll();
    tasksFinished.await();

    // Check all of the removal notifications we received: they should have had correctly-associated
    // keys and values. (An earlier bug saw removal notifications for in-progress computations,
    // which had real keys with null values.)

    int numActions = 100;
    // hashcodes from non-overlapping hash computations
    HashCode expected1 = randomHash(hashFunction, new Random(1L), numActions);
    HashCode expected2 = randomHash(hashFunction, new Random(2L), numActions);

    // equivalent, but overlapping, computations (should produce the same results as above)
    Random random1 = new Random(1L);
    Random random2 = new Random(2L);

   * usually have much noticeable impact. A value of one permits only one thread to modify the cache
   * at a time, but since read operations and cache loading computations can proceed concurrently,
   * this still yields higher concurrency than full synchronization.
   *
   * <p>Defaults to 4. <b>Note:</b>The default may change in the future. If you care about this

      }
    }.start();

    getStartedSignal.await();

    // computation is in progress; asMap shouldn't have changed
    assertEquals(1, map.size());
    assertFalse(map.containsKey(getKey));
    assertSame(refreshKey, map.get(refreshKey));

    // let computation complete
    letGetFinishSignal.countDown();
    getFinishedSignal.await();
    checkNothingLogged();

.concat([fragments[divider].split(splitRegex)[0]]),\n        [fragments[divider].split(splitRegex)[1]].concat(\n          fragments.slice(divider + 1)\n        ),\n      ]\n    : [fragments];\n\n  // Convert the values with units to absolute pixels to allow our computations\n  ops = ops.map((op, index) => {\n    // Most of the units rely on the orientation of the popper\n    const measurement = (index === 1 ? !useHeight : useHeight)\n      ? 'height'\n      : 'width';\n    let mergeWithPrevious = false;\n    return...

   * usually have much noticeable impact. A value of one permits only one thread to modify the cache
   * at a time, but since read operations and cache loading computations can proceed concurrently,
   * this still yields higher concurrency than full synchronization.
   *
   * <p>Defaults to 4. <b>Note:</b>The default may change in the future. If you care about this

  * Note that NumPy's type promotion rules have been changed(See [NEP 50](https://numpy.org/neps/nep-0050-scalar-promotion.html#nep50)for details). This may change the precision at which computations happen, leading either to type errors or to numerical changes to results.

   * The listener will run when the {@code Future}'s computation is {@linkplain Future#isDone()
   * complete} or, if the computation is already complete, immediately.
   *
   * <p>There is no guaranteed ordering of execution of listeners, but any listener added through
   * this method is guaranteed to be called once the computation is complete.
   *