public static int divide(int dividend, int divisor) {
    return (int) (toLong(dividend) / toLong(divisor));
  }

  /**
   * Returns dividend % divisor, where the dividend and divisor are treated as unsigned 32-bit
   * quantities.
   *
   * <p><b>Java 8+ users:</b> use {@link Integer#remainderUnsigned(int, int)} instead.
   *
   * @param dividend the dividend (numerator)
   * @param divisor the divisor (denominator)

 x     y     x / y     x % y
 5     3       1         2
-5     3      -1        -2
 5    -3      -1         2
-5    -3       1        -2
</pre>

<p>
The one exception to this rule is that if the dividend <code>x</code> is
the most negative value for the int type of <code>x</code>, the quotient
<code>q = x / -1</code> is equal to <code>x</code> (and <code>r = 0</code>)

import static com.google.common.base.Preconditions.checkPositionIndexes;
import static com.google.common.base.Preconditions.checkState;
import static com.google.common.math.IntMath.divide;
import static com.google.common.math.IntMath.log2;
import static java.lang.Math.max;
import static java.lang.Math.min;
import static java.math.RoundingMode.CEILING;
import static java.math.RoundingMode.FLOOR;

      // Avoid delegating to this(long[]), since AtomicLongArray(long[]) will clone its input and
      // thus double memory usage.
      this.data =
          new AtomicLongArray(Ints.checkedCast(LongMath.divide(bits, 64, RoundingMode.CEILING)));
      this.bitCount = LongAddables.create();
    }

    // Used by serialization
    LockFreeBitArray(long[] data) {
      checkArgument(data.length > 0, "data length is zero!");

      // non-negative int, we can do long-arithmetic on index * (dataset.length - 1) / scale to get
      // a rounded ratio and a remainder which can be expressed as ints, without risk of overflow:
      int quotient = (int) LongMath.divide(numerator, scale, RoundingMode.DOWN);
      int remainder = (int) (numerator - (long) quotient * scale);
      selectInPlace(quotient, dataset, 0, dataset.length - 1);
      if (remainder == 0) {

  public void testBitSize() {
    double fpp = 0.03;
    for (int i = 1; i < 10000; i++) {
      long numBits = BloomFilter.optimalNumOfBits(i, fpp);
      int arraySize = Ints.checkedCast(LongMath.divide(numBits, 64, RoundingMode.CEILING));
      assertEquals(
          arraySize * Long.SIZE, BloomFilter.create(unencodedCharsFunnel(), i, fpp).bitSize());
    }
  }

  /**

	VMULHSUVV	V1, V2, V3			// d7a1209a
	VMULHSUVV	V1, V2, V0, V3			// d7a12098
	VMULHSUVX	X10, V2, V3			// d761259a
	VMULHSUVX	X10, V2, V0, V3			// d7612598

	// 31.11.11: Vector Integer Divide Instructions
	VDIVUVV		V1, V2, V3			// d7a12082
	VDIVUVV		V1, V2, V0, V3			// d7a12080
	VDIVUVX		X10, V2, V3			// d7612582
	VDIVUVX		X10, V2, V0, V3			// d7612580
	VDIVVV		V1, V2, V3			// d7a12086

        passed to the optimizer but the reported loss will be a scalar value.
    *   `SUM`: Scalar sum of weighted losses. 4. `SUM_OVER_BATCH_SIZE`: Scalar
        `SUM` divided by number of elements in losses. This reduction type is
        not supported when used with `tf.distribute.Strategy` outside of
        built-in training loops like `tf.keras` `compile`/`fit`.