Thread: Floats & memory

I know all about integer overflows, but not of floating point overflows. I noticed something odd today. Apparently I can store whatever number I want inside MoHAA's floating point datatype.
It became evident today when I used the new pow function with the arguments 2 & 32. I was under the impression that floats were 32 bits long much like the integers in MorpheusScript but
pow 2 64 was successfully stored in my variable and printed the exact value (except for the first 3 digits, they were 0).

It also works if I directly define a large float:

Code:

 local.var = 4294967296.0 //2 ^32

, which should overflow because both integers and floats in MoH are signed.
The greatest value I could largely fit in there was "170141183460469230000000000000000000000.000", after that it kept returning "1.#IO".
The thing is while I can't perform addition or subtraction with these values, I can still correctly divide, multiply and compare them, which makes this 'trick' very interesting to me,
especially since 2^32 is a value that can be quite useful.

Can anyone explain why and/or how this is happening, and if this can cause any kind of stability issues or memory leaks?

Because floats are not stored as a concrete value really.

You can read about it here: http://en.wikipedia.org/wiki/Floatin...dern_computers

Take a look at internal representation section. It's a bit like scientific notation.

Any integer with absolute value less than or equal to 2²⁴ can be exactly represented in the single precision format, and any integer with absolute value less than or equal to 2⁵³ can be exactly represented in the double precision format. Furthermore, a wide range of powers of 2 times such a number can be represented.

Hah, how about that, I did in fact only use multitudes and powers of 2 to test it.

This means we can also use the infinities in scripts:

Code:

 level.MAX_INFINITY = pow 2 128 // => "1.#IO"
level.MIN_INFINITY = pow -2 129          // => "-1.#IO"