Hi,

I am wondering if anyone has "expert tips" when doing very high resolution renderings in Processing ...

For instance, on a Mac Pro laptop with 4 GB of RAM, what is the optimum memory size to choose?

I want to do the initial rendering at 3X or 4X the final target resolution so I can do some Gaussian filtering and down-sampling to overcome the moire' pattern and aliasing that result from many fine lines close together.

Incidentally, the default Processing renderer (is it P2D, or JAVA2D?) gives visually better results than the OPENGL renderer.  Are there some OPENGL tricks I could be using to speed things up and/or get better anti-aliasing performance?
Would rendering directly to PDF help with the quality and/or speed?  There are many many many curves drawn.  I don't know if that would pose a problem for PDF.

Rendering time is also a concern:  When my target resolution gets beyond 5000x4000, the rendering really starts to bog down.  I was hoping I could get closer to 9000x9000, but that seems impractical right now.

By the way, my Processing sketch looks like a translucent silk scarf.

http://www.flickr.com/photos/dxjones

If anyone has practical experience with very high resolution renderings (for producing large printed images), I would love to learn what helpful tips you have discovered along the way.

thanks in advance,
 David Jones

I took a quick look at "Tile Saver".  Unfortunately, it requires the entire scene to be re-rendered for each tile.  This would take far too long for me, ... plus the added complexity of restoring the same initial conditions for the random number generator to retrace the exact same steps in the long sequential rendering computation.

As for my technique, ... yes you are on the right track.  I use a chain of particles, physics, bezier curves, perlin noise.

By the way, one of my frustrations is that Processing (currently) implements the "old" version of Perlin Noise, which suffers from visibly noticeable "second order discontinuities" at integer boundaries.  This defect is particularly ugly when using noise() to control movement.

Ken Perlin published his "Improved Noise" paper at SIGGRAPH 2002, and his reference implementation in JAVA is available online:
http://mrl.nyu.edu/~perlin/noise/

I wish we could "upgrade" the Processing noise/noiseDetail functions to incorporate this "new and improved" code.  I don't know what would be involved in doing this.  I realize Perlin's code can't simply be dropped in, since there is an extra layer of "noiseDetail".  Does anybody know who write the version noise/noiseDetail currently in Processing?  Perhaps they could do the upgrade.

-- djones

The current version of perlin noise in Processing "works", but has a defect when parameters cross integer boundaries, where it becomes less "smooth".  Ken Perlin has fixed the problem with "improved noise", and he even provides Java code.

If I get some free time, I will post a simple applet that illustrates the phenomenon.  I looked at the noise code inside Processing and it looks like it was written by "toxi".  Perhaps he might be persuaded to do the update.

-- djones

Hmm, did you take a look at his work?
Would you try to to do it with a pdf ?
Dontk now how much lines there are but looks like alot. Only thing i can say, i produced pdfs arround 250mb and it was absolutly impossible to open them withought crashing Acrobat, Illustator and so one.
But i would be thankfull for any hint

Ah that sounds good. Maybe it looked more complex than it was. How big are your PDF files getting at most ?
thats how most of my images looked like : http://projekttriangle.com/images/content/news/images/Burda_news.jpg, was just not possible with pdf export.

but like you say, its important to get a good workflow, cause testing by rendering and exporting and opening in photoshop to check cost alot of time.

Link seems to be broken... anyway. rendering was never the problem. I always had problems to open larger pdf files bigger then 30-50mb. always crashed photoshop/illustrator.