Convert float to int bitwise without casting

Question

I'm trying to checksum an array of floats. Therefore I want to convert them to integers bitwise. Is there any fast way to this not manipulating the bytes by myself?

Something like [`Float.floatToRawIntBits()`?](https://docs.oracle.com/javase/8/docs/api/java/lang/Float.html#floatToRawIntBits-float-) — Andreas Fester, Apr 13 '16 at 18:07
Yes, thanks! Actually, what is the difference between floatToIntBits() and floatToRawIntBits()? The handling of NaN? — Geosearchef, Apr 13 '16 at 18:11
Yes, if the actual number is representing a `NaN` value then the two methods return different values. `floatToIntBits()` returns the same canonical value for all possible `NaN` values while `floatToRawIntBits()` returns different values, depending on the actual `float` `NaN` value. — Andreas Fester, Apr 13 '16 at 18:15

score 6 · Accepted Answer · edited May 23 '17 at 12:31

You can get the internal, bitwise representation of a float with Float.floatToRawIntBits(). You can then use the returned int to create a checksum:

int checksum = 0;
for (float value : floatArray) {
  checksum += Float.floatToRawIntBits(value);
}

Depending on your use case, you might also use Float.floatToIntBits() which returns the canonical value 0x7fc00000 for all NaN values.

As for the checksum calculation itself, I took the simplest approach of just summing up the values. There are better checksum algorithms which are more robust against specific flipped bit patterns - once you have the int values, you can use them with any checksum algorithm you like (CRC, MD5, SHA2, or something else).

For more information, see for example

I suggest using a multiplier to create a simple hash. But this does the job +1 — Peter Lawrey, Apr 13 '16 at 18:37

Convert float to int bitwise without casting

1 Answers1