OT : Approximate / fast math libraries ?
Chris Jones
jonesc at hep.phy.cam.ac.uk
Sun Sep 2 19:46:23 UTC 2007
Hi,
> You could try this piece of code:
<snip>
Many thanks, such a power series solution was what I was hoping to find in
some standard library. However, reading around now I've come to the
conclusion that the acceptable error in cases where approximate solutions are
valid naturally varies from project to project, so its not really possible to
write an optimal approximate solution that fits all.
I've come to the conclusion that I need to do this myself. And here I thank
you for your code, which I've taken as a bas is for a solution. many thanks.
Tests indicate it does lead to a performance gain which similar to what you
quote below.
>
> atan2: 7.028u
> atan2x: 1.586u
>
> I'm not an expert on this, so use it at your own risk :-)
no worries ;)
cheers Chris
More information about the fedora-list
mailing list