for z := 0; z < REP; z++ {
		// pick a random key, hash it
		k.random(r)
		h := k.hash()

		// flip each bit, hash & compare the results
		for i := 0; i < n; i++ {
			k.flipBit(i)
			d := h ^ k.hash()
			k.flipBit(i)

			// record the effects of that bit flip
			g := &grid[i]
			for j := 0; j < hashSize; j++ {
				g[j] += int(d & 1)
				d >>= 1
			}
		}
	}

	for z := 0; z < REP; z++ {
		// pick a random key, hash it
		k.random(r)
		h := k.hash()

		// flip each bit, hash & compare the results
		for i := 0; i < n; i++ {
			k.flipBit(i)
			d := h ^ k.hash()
			k.flipBit(i)

			// record the effects of that bit flip
			g := &grid[i]
			for j := 0; j < hashSize; j++ {
				g[j] += int(d & 1)
				d >>= 1
			}
		}
	}

        `FlushCaches()`.
    *   Add `auto_correlation` to distributions.
    *   Add `tf.contrib.distributions.Autoregressive`.
    *   Add SeparableConv1D layer.
    *   Add convolutional Flipout layers.
    *   When both inputs of `tf.matmul` are bfloat16, it returns bfloat16,
        instead of float32.
    *   Added `tf.contrib.image.connected_components`.