PDA

View Full Version : Secret 2011 optimization?


destruct007
04-18-2010, 02:55 PM
Secret in that I never heard anyone talk about it, Hermite interpolation is no longer slow!?! :bounce: Its still slower than linear but its definitely more workable. Wow, now in addition to learning all the new attrs I have to re evaluate what I already use if hermite is better.

Is it just me or is the solver all around faster? It sure feels faster.

viki164
04-19-2010, 09:21 AM
In 3d container its still dog slow, but I agree with 2d the speed looks more faster then it was in maya 2010.
btw since Forward advection & substep is introduced then I dont really find the hermite solver of much use. whats your take dave?
Do people really make a good use of hermite?

YourDaftPunk
04-19-2010, 05:45 PM
Hermite was never fast enough to use in production, but low res tests in 2009 showed more detail than linear.

Interpolation examples:
http://local.wasp.uwa.edu.au/~pbourke/miscellaneous/interpolation/ (http://local.wasp.uwa.edu.au/%7Epbourke/miscellaneous/interpolation/)

Why doesn't maya have a cubic method as a compromise between linear and hermite? The hermite example has tension and bias values to adjust the interpolation, but we don't have these controls exposed in maya. And it takes longer- look at those extra operations.

KidderD
04-19-2010, 08:07 PM
Talk is cheap... let's see some examples. :)

Aikiman
04-19-2010, 08:08 PM
Yeah I tried that one too but still really slow and waste of space. Switch to self attract and repulse and thats slo enough for me. But then my hobby mac has been called a "shopping cart" at an Indie 500.

Aikiman
04-19-2010, 09:30 PM
Hermite was never fast enough to use in production, but low res tests in 2009 showed more detail than linear.

Interpolation examples:
http://local.wasp.uwa.edu.au/~pbourke/miscellaneous/interpolation/ (http://local.wasp.uwa.edu.au/%7Epbourke/miscellaneous/interpolation/)

Why doesn't maya have a cubic method as a compromise between linear and hermite? The hermite example has tension and bias values to adjust the interpolation, but we don't have these controls exposed in maya. And it takes longer- look at those extra operations.

Cubic looks just as long bar one or two lines of code, perhaps that makes all the difference when we are talking millions of calculations per frame step. Not sure why you couldnt use cosine though?!

Duncan
04-19-2010, 09:58 PM
Using cubic interpolation in the solve (as opposed to the rendering) just results in huge amounts of diffusion and thus was not included as an option.

Duncan

YourDaftPunk
04-20-2010, 02:46 PM
Using cubic interpolation in the solve (as opposed to the rendering) just results in huge amounts of diffusion and thus was not included as an option.

Bummer. I wish I knew more of the math involved. I thought I had seen cubic as an option in another fluid solver (it was likely rendering related interpolation then). Thanks for the answer Duncan.

-shawn

tpalamar
04-20-2010, 04:12 PM
Hermite definitely adds more detail. Looks a lot better, but did take about four times longer to process. Here is a comparison.


http://www.speffects.com/images/postings/linear_vs_hermite.jpg

destruct007
04-21-2010, 12:37 AM
I did tests with hermite but there's so much else to get used to and learn, and re-find my happy settings that I left it alone. I didn't get anything worth showing, but it is something I will go back to.

jedipottsy
04-23-2010, 08:03 PM
I downloaded the trial and im really impressed. I really like that particles can emit fluids now, this is a quick test i made

http://mythreed.com/videos/MagicDust.mov

CGTalk Moderation
04-23-2010, 08:03 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.