static int rehash(int h) {
    // Spread bits to regularize both segment and index locations,
    // using variant of single-word Wang/Jenkins hash.
    // TODO(kevinb): use Hashing/move this to Hashing?
    h += (h << 15) ^ 0xffffcd7d;
    h ^= h >>> 10;
    h += h << 3;
    h ^= h >>> 6;
    h += (h << 2) + (h << 14);
    return h ^ (h >>> 16);
  }

  /**

- cloud.google.com/go/gkeconnect: v0.8.4
- cloud.google.com/go/gkehub: v0.14.4
- cloud.google.com/go/gkemulticloud: v1.1.0
- cloud.google.com/go/gsuiteaddons: v1.6.4
- cloud.google.com/go/iam: v1.1.5
- cloud.google.com/go/iap: v1.9.3
- cloud.google.com/go/ids: v1.4.4
- cloud.google.com/go/iot: v1.7.4
- cloud.google.com/go/kms: v1.15.5
- cloud.google.com/go/language: v1.12.2

[<a href="#Go_1.13">Go 1.13</a>].
A single <code>0</code> is considered a decimal zero.
In hexadecimal literals, letters <code>a</code> through <code>f</code>
and <code>A</code> through <code>F</code> represent values 10 through 15.
</p>

<p>
For readability, an underscore character <code>_</code> may appear after
a base prefix or between successive digits; such underscores do not change
the literal's value.
</p>
<pre class="ebnf">

    *   Support loading of custom operations (e.g. many in `tf.contrib`) on
        Linux and OS X
*   All our prebuilt binaries have been built with CUDA 8 and cuDNN 6. We
    anticipate releasing TensorFlow 1.5 with CUDA 9 and cuDNN 7.

## Bug Fixes and Other Changes

*   `tf.nn.rnn_cell.DropoutWrapper` is now more careful about dropping out LSTM