I'm trying to use float in 16bit environment. But I can't figure out how to declare an single precision float.

This is what I'm trying to do.


The result is: dx:ax = 1|11110001|100011101000011001000 = Very large negative number.

[edit]
Are all declared float treated as double precision?
[/edit]

Hehe, if you read my previous post you can atleast see one error in it. I'm trying to represent a double word with 30bits.
What I have learned from this is to never trust Microsoft's calculator as it cuts of ending zeros. The bits should look like this:

0|01111100|01100011101000011001000

If I have understood floating numbers this is how it works. Lets say we have bit 0 to the right and bit 31 to the left.
Bit31 = 1 if negative otherwise 0
Bit23-30 = Exponent
Bit0-22(23) = Mantissa

Bit23 is a very special bit since it being assumed to be 1 when used together with mantissa even if it is 0 in my example above. It is only assumed to be 0 in special cases. The mantissa is represented like this when you calculate it:
1.01100011101000011001000
2^0+0^-1+2^-2...+0^-23
If you want to make it easy on yourself you put everything after the dot in a calculator converts it to decimal and divide it by 2^23 and then add 1.

Conclusion:
What I tried to represent as a dw was 0.173648 if you do the calculation from my bits above you get. 1*2^(124-127)*1.389183=0.1736478. So it seems my previous code was right and Fasm handles my Sin value as single precision.

If you would like to know more about float go here http://stevehollasch.com/cgindex/coding/ieeefloat.html

Hopefully this helps some confused float newbies as myself.