#58889 - strager - Wed Oct 26, 2005 11:57 pm
Does anybody know how to handle powers/exponents with fixed numbers?
Here's the actual situation: I have two fixed numbers, both 16.16, and I need to raise one to the other; as in such:
a ^ (1 / 2.4)
a is some random number calculated by the parameters to the function and other factors. I need to know how to calculate that in C... because using floats is slow 'n' stuffs...
Thanks!
#58898 - tepples - Thu Oct 27, 2005 12:38 am
What will you use this value for? If it's for gamma correction, then use a lookup table.
_________________
-- Where is he?
-- Who?
-- You know, the human.
-- I think he moved to Tilwick.
#58901 - strager - Thu Oct 27, 2005 1:12 am
| tepples wrote: |
| What will you use this value for? If it's for gamma correction, then use a lookup table. |
Basically, but how am I going to use a lookup table for all possible 4 billion values? Maybe I could leave only the 8 most accurate bits and do a LUT on that? Or what would you recommend I do?
Thanks...
#58903 - tepples - Thu Oct 27, 2005 1:31 am
| strager wrote: |
| but how am I going to use a lookup table for all possible 4 billion values? Maybe I could leave only the 8 most accurate bits and do a LUT on that? |
Remember that all you need is a 5-bit output per channel. Use enough of a lookup table that gives you accuracy down to 5 bits.
_________________
-- Where is he?
-- Who?
-- You know, the human.
-- I think he moved to Tilwick.
#58915 - SevenString - Thu Oct 27, 2005 5:44 am
As a shortcut, sometimes linear interpolation between lookup table values can provide reasonable results, even if the lookup table is non-linear.
_________________
"Artificial Intelligence is no match for natural stupidity."
#58965 - legalize - Thu Oct 27, 2005 5:50 pm
Have you tried looking at the sources to a free math library, like oslib on sourceforge?