2
$\begingroup$

0.1 via $32\text{ bit}$ is rather easy:

Sign: $0_2 = 0_{10}$

Exponent: $123_{10} = 01111011_2$

Mantissa: $5033165_{10} = 100110011001100110011001101_2$

Now, how do you calculate this, if you've only $8\text{ bit}$ available?
Sign: $1\text{ bit}$, Exponent: $2\text{ bit}$, Mantissa: $5\text{ bit}$

My idea: $S=0; E=01; M=00001$

What do you think?

$\endgroup$
  • $\begingroup$ There is no 754 standard for 8 bit mini-floats. $\endgroup$ – Ranbir Feb 1 '15 at 19:38
2
$\begingroup$

For $8$-bit floating point, you should look at $A$-law and $\mu$-law encoding standards:

http://en.wikipedia.org/wiki/%CE%9C-law_algorithm

http://en.wikipedia.org/wiki/A-law_algorithm

Depending on your input range, you may need some scaling.

$\endgroup$

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.