enter image description here

Hi, please help me on this problem, I don't understand how this Float "4-bytes" or IEEE 754 floating point, is converted to short_signed "2-bytes".

First, I convert the float "4-bytes" hexadecimal to float decimal, then I convert this float decimal to hexadecimal again. and I get the short_signed result.

But the problem when I put the normals or UVs, I get always 0.xxxx when I convert this 0.xxxx to hexadecimal doesn't give the result like the short_signed in the image or excel file.

Here is the website to convert hexadecimal to IEEE 754: https://gregstoll.com/~gregstoll/floattohex/ (check swap endianness) and the website to convert decimal to hexadecimal: https://www.rapidtables.com/convert/number/decimal-to-hex.html

Here is the full Float 4-bytes to Short_signed 2-bytes excel file for results: https://drive.google.com/open?id=1oKxVuqFVcVX3NEGLwzFvsZqXos1PwYIA

Thanks in advance.

  • 1
    I might be missing the point but how does it relate to RE? – Paweł Łukasik May 12 '20 at 15:41
  • "First, I convert the float "4-bytes" hexadecimal to float decimal, then I convert this float decimal to hexadecimal again. and I get the short_signed result." That makes no sense. Add how exactly you are 'converting'. – Jongware May 12 '20 at 21:26
  • 1- Copy "4-bytes" hex from excel file. 2- Paste it into hex value in: gregstoll.com/~gregstoll/floattohex 3- Check swap endianness. 4- Click on convert to float. 5- Copy the float value. 6- Paste it into decimal number in: rapidtables.com/convert/number/decimal-to-hex.html 7- Click on convert. 8- And you get the "2-bytes" hex like in excel file. – Segal May 13 '20 at 13:29

One word, precision

By definition, a normal is a vector with a length of 1.0f which means that each component of the vector will always fall between -1.0f and +1.0f, and UV are texture coordinates expressed in a range between 0 and +1.0f.

The number of possible states that you can represent with this range of numbers is huge and even exceeds the total number of states in a singed short. So if you want to retain the greatest precision possible, you need to scale the value by the maximum value of a singed short.

Therefore, to obtain the correct results, you need to multiply by 32767 and then round it to the nearest integer (if it's a texture coordinate) or truncate it (if it's a normal). The rounding and truncating are just some observations I made when converting the values myself.

Example normal (using the 9th row):
X: 0x96DE = -26914 = truncate(-0.821400702 * 32767 = truncate(-26914.836802434)
Y: 0x3E17 = 13847 = truncate(0.422606796026 * 32767) = truncate(13847.556885383942)
Z: 0xCEFA = -12250 = truncate(-0.383020102978 * 32767) = truncate(-12550.419714280126)

Example UV (using the 1st row):
X: 0x64FD = 25853 = round(0.788999974728 * 32767) = round(25853.162171912376)
Y: 0x5D49 = 23881 = round(0.72879999876 * 32767) = round(23880.58955936892)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.