Reducing Convolution Filter Rendering Time
From Ultrafractal Wiki
also see: Rendering Time, Convolution Filtering, Image Traps Tweak
Connie said:
this one is a long render so i made it smaller and got tired of trying to rerender for a better composition. its been rendering to disc for 16 hours in a larger size on my computer and has 9 hours to go. i wanted to try the convolution wrapper, sooooo, long render.
Connie
(db: for Connie's Challenge9+cld2b upr see: Image Traps Tweak )
***
Hi Connie (and hopefully useful advice to the rest of you),
Concerning your challenge9 entry, the Convolution Filters are very time demanding, as you found. I would recommend limiting convolution to one layer only, and for testing/optimizing purposes, to set the sample density to 1. The convolution filter samples #z values around the displayed value to calculate the convolution. with samples = 1, there are 9 #z values sampled. With samples = 2, 25 #z values are sampled, etc.
If you want the filters on all three layers I would advise optimizing your image using the "Null Filter" and then adding the filters you want back in, first with samples = 1, and then perhaps samples = 2.
You are using an old version of image traps on your top layer, but not on the next three layers.
On all of your layers you have the "Default Color Merge Modes" set to "normal" and "image on top". For those settings the coloring in the gradient layer are basically not being used, and so you lose all the shading and coloring effects that come from the gradient layer. Is that your intent?
***
More advice and comments to everyone:
We have had definite interest in ufm choices and related issues on the list. After playing some more with Connie's entry I found that much of the speed of rendering problem relates to other choices in the layers rather than the convolution filter. I replaced the convolution filter with the null filter on all layers, and rendering was still VERY slow. I then observerd that two layers had both inside and outside coloring. Inside coloring with a relative high iteration value tends to be very slow. My final observation relates to the ufm itself. Phoenix is one of my favorite ufms, but as a formula, it also runs slower than many other formulas, although this was the least important contributer to the rendering speed.
