| ??? 06/10/02 22:49 Read: times |
#24242 - RE: floating point optimization |
If speed is _really_ important, use assembly. Much faster, if you exactly know, what precision you really get from the input and what precision you really need at the output. Using assembly, you can use exactly the necessary precision for the data, different precision for different internal data calculations, tricks, precalculated tables etc. Not easy, but of course much more efficient then c libs. |



