Filter Forge - Weak floating point precision.

Active Topics Search

Andrey Mironov Andrey Mironov Posts: 39 Filters: 36	This filter shows FF's weak floating point precision. I am VERY sorry, BUT according to all elementary math rules: this cascade of differeces between sequental images of differecnes between (...and so on) MUST converge to ZERO with ANY exponent value (reversed Pascal's triangle). It IS so with exp value of 1. Other exp value makes FF glitch and comes up with an error of about 0.01 (!!!) and higher. I believe, precision leaks out within EXP math operation... or accumulates within every difference blending call. Be careful doing your math! Filter: Result with exp = 2: WeakPrecision.ffxml Andrey Mironov
	Posted: December 3, 2015 2:43 pm	Details E-Mail
xirja Idididoll Forcabbage Posts: 1698 Filters: 8	I haven't checked the exp function specifically, but on the contrary generally, at least in version 4, the following wouldn't have worked if the precision was off. https://www.filterforge.com/forum/read...&TID=11972 TIFF, OpenEXR, and PFM (32 bit float) formats may help. The eye dropper and the input fields don't have much precision (visibly), but ffxml file shows otherwise. _____________________________________________________ http://web.archive.org/web/2021062908...rjadesign/ _____________________________________________________
	Posted: December 3, 2015 4:26 pm	Details E-Mail

Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!

33,711 Registered Users
+18 new in 30 days!

153,531 Posts
+36 new in 30 days!

15,347 Topics
+72 new in year!

16 unregistered users.