hornet777

2nd November 2005 07:26 UTC

numerical limits, et al
I had three questions; I hope they aren't redundant, as I used search before posting this, and found nothing.

First, what are the numerical limits of AVS? that is, in any given function, what are the largest and smallest numbers it will recognise? E.g., in superscope, how far could one go in setting "n"? Similarly, what is the smallest (floating-point) number that can be recognised? This is really a question about AVS's precision.

Second, why hasn't scientific notation nomenclature been implemented in AVS? This alone could reduce some of the bloat of entering huge/small numbers, and make things a bit easier to read. (E.g., 1E06 for one million)

Lastly, when can we expect a release at the sourceforge site for a new version of AVS? I know the latest is included with the latest Winamp, but if I'm happy with what I have (next to last) why should I install something I don't necessarily want just to get the latest version of only part of it when its supposedly already "open source"? (This is not a bitch or rant.)

I really love AVS, playing with it, and learning more about its obscurities and intricacies (and Winamp too). Many thanks for all those who have made it/them available.

gordmoo

2nd November 2005 09:36 UTC

avs uses 64-bit floating points if i remeber right. so your basically looking at 17 or so digits.. allthough i might be out by a bit there.

i guess scientific notation hasn't been encorporated in because the evalib language isn't exsacly a compleate language, n so its not needed really.

n as far as i know, there is no scedualed releace for the sourceforge site, if anything fun happens or avs does actually update one day, im sure it will be reflected in there. the version you get with winamp with prolly be the latest version, other than that just check back once in a while.

Warrior of the Light

2nd November 2005 13:58 UTC

you could easily test the max of n by creating a solid dotgrid with n=w*h at the highest posible res and see where it stops.

But with n=w*h, the preset will be too slow to watch anyway, so it's really useless to know.. It's sufficient.

hornet777

3rd November 2005 05:49 UTC

Thanks for your responses. Dumb ass me, I forgot to say that not only was I interested in (say) n, but how may places $PI, or $E or $PHI was calculated to... I assume the 64-bit, 17 place answer above would apply. It wasn't so much to explore thse outer reaches in practice, but to determine how much of an advantage there was in using the preset variable as opposed to typing in a few places.

Learning a lot, and having a ball.

Added a bit later:

Also i just found an old post from UnConed as a rough guide to when high precision/resolution is appropriate.

gordmoo

5th November 2005 20:21 UTC

just as a note, if you type in a number, no matter how large it is. it will be shortened to the double float.

hornet777

7th November 2005 07:05 UTC

cool, thanks....very helpful