Welcome to Fractal Forums

Kalles Fraktaler is really a collection of all the amazing things discovered on this forum :)

I have today uploaded a new version of Kalles Fraktaler, version 2.9.3
Get it from http://www.chillheimer.de/kallesfraktaler

There is only one new function added, the Newton-Raphson method of finding minibrots or zoom to 3/4 of the minibrot, where the pattern of the current view is doubled.
Thanks to claude for the theory and source code!

Julia morphing has never been easier.
Zoom to an interesting spot.
Hit Ctrl+H to pre-view inflection, i.e. how the view will look 3/4 to the minibrot where the pattern is doubled.
Hit Ctrl+D and press the left mouse button.
Wait for the magic to be complete, go grab a coffe, go get a beer at the pub, socialize with your fiance or sleep for the night, and let the computer do the dirty work.
(you may need to do some fine tuning zooming in advance or afterwards, or the function will fail, but it gets more reliable the deeper you go, and more time consuming...)

A test image is attached, zoom depth of 5.26e128 made in minutes.

I've tried it be somehow did not work. Or I just don't have the right version. Has the 32bit version been updated as well?

(http://i.imgur.com/KcTfFjq.jpg)

Sorry this update was only the 64 bit version.
Do you want it also in 32-bit? I thought you had got yourself a 64-bit machine now? :)

That would be great.

Yes, I do have a 64bit machine but it is dusty and although I can do a bit of cleaning it is a huge case  :-\. The images made with 64bit I mentioned were on my fathers laptop. That's also why the last two panorama's were made with a 32bit machine.

32-bit version of 2.9.3 now uploaded. :)
I was not able to compile it in VS6 so I hope you can solve any eventual problems with missing dlls (msvcr100.dll)

I haven't had any problems so far and the morphing works perfectly. I will have to get used to the fact that I need a little bit more precise when clicking but it is a nice feature. Thanks for adding the texture feature as well. :)

Version 2.9.4 uploaded, with some improvements in the Newton-raphson method

I'm running into stability issues when I try to make very high-resolution images.

I make very large fractal prints. In order to anti-alias them for printing, I render at 4x the final size. This requires some very high resolutions. Some of my images have been as big as 30,000 pixels across.

If I try to render images in KF bigger than about 8kpx, I run in to extreme instability. Sometimes KF stalls without drawing anything, or using any CPU. Sometimes it throws hundreds of error messages. Sometimes it crashes and the OS notifies me.

My render stations run Windows 10 and 7. The W10 box is a hex-core Xeon X5680 and my W7 box is a i7-4790k. The Xeon has 24 GB of RAM and the i7 has 32GB. I never see more than half used on the Xeon, so I know it's not running out of RAM.

I'd like to render arbitrarily large, multi-gigapixel images.

Thank you.

Thanks for you comment!
Are you using the 64-bit version of Kalles Fraktaler?

Yes, 64 bit on both boxes.

I think programs that works with memory bitmaps are limited to a half giga pixels
http://www.fractalforums.com/index.php?topic=23336.msg91165#msg91165
However in Kalles fraktaler you can move the view exactly one window up, down, left or right with ctrl+arrow keys. Then you can store several images and "stitch" them together with some other program that doesn't work with full bitmaps (photoshop?)
I think TheRedshiftRider does this when he makes his giga pane images?

I use ImageMagick to do re-sizing of large images. For example, I generate a 30,000 x 20,000 pixel image in Fractal Extreme, then resize it using ImageMagick command-line tools. I then send the final image to the online printing service, and they mail me a 300dpi print at about 24 x 16 inches.

I'd like to make even bigger images (64,000 pixels) using KF, but it crashes.

Thank you.

I am currently rendering a zoom movie with Kalles Fraktaler. I wanted to render the images at 2560x1440 to do an anti-aliased video at 1280x720. When it asked me to export the JPG images, I specified 1280x720 size expecting anti-alieased previews of what the video would look like. I only have 146Gb free space on the hard drive partition I am using to store the zoom sequence. Anyway my movie goes down to 3975 zoom levels and I estimated with 2560x1440 frames being stored as 28.1 Mb a piece, they would take up to 118Gb on my hard drive, plus the preview JPGs. I didn't want to run out of space due to the JPGs so I specified 1280x720 for the sample JPG resolution. Now it appears that Kalles Fraktaler is processing the frames at 1280x720 resolution, which means I will not be getting an anti-aliased video. I am barely e10 down the zoom right now from the Minibrot. It's mostly dense circles at this point anyway, and since the last 5% of the zoom movie takes the bulk of the render time, would it be safe if I manually change the image resolution to 2560x1440 in KF so that only the last few seconds of the resultant video will not be antialiased? I don't want to get corrupted data when I try to render video and it's only 18 hours in but I'd rather not start over. If I render the remaining frames at 2560x1440, will it corrupt the existing sequence? Thanks...

I have done such operations myself.
If you change the resolution when rendering, the render will stop and an ordinary render will be processed, which you may break.
Just start the zoom sequence again, from that position, to the same directory, the image counter will resume the sequence, the numbering will be correct.
KeyFramMovieMaker is able to handle frames with different sizes.

Okay so Kalles Fraktaler has stopped as you predicted. I just resave zoom sequence into the same directory?

EDIT: You were right. It picked up right where it left off. Perhaps I should have let it run longer at lower resolution? It will be slow go for a while, but as we all know with zoom movies, the last ten percent (or first in the case of reverse order rendering) takes up 90% of the render time. I will wake up some morning and it will be done. Also I need to invest in a bigger hard drive. KF, have you thought about adding a compression algorithm to the zoom movie frames? The source files are quite large but I would like to store them in case I want to re-render he in the future.

I predict I'll have about 10-15 Gigs left on my "Stuff" partition when I'm done rendering this movie. My EATX YY201 workstation chassis has a fan cooled hard drive bay in the front with space for up to five hard disks. Copying files across multiple drives isn't enough. I should probably also get some offline external storage in case  my PC ever gets hit by ransomware, so I won't lose everything... :fear: :death:

Suggestion: Accept a file to open as the first command line argument

I'm having fun with this!  Even though it does get slow after a few goes around, and I had one instance when KF stopped responding with the find period counter stuck on 880000 or so... had to kill it, didn't get around to retrying yet.

Feature request: extend it to arbitrary-power Mandelbrot (power k needs to iterate z→z^k+c for box-period corners, and Newton method derivative needs to be dz→k dz^(k-1)+1, other than that most of the code should remain the same).  But: I don't know if the size formula will work for higher powers (with iterations of z replaced appropriately), maybe it would work enough to guess the depth?

I have found the higher order mandelbrots tend to hit the minibrot at a much shallower depth. At larger orders (6 and higher) it becomes blobby and you have to zoom quite deep just to get away from visible minis.

For instance, from past observations:

Second Order:
Pattern (forked path) = 1x depth, 1x iterations
Pattern doubled = ~1.5x (3/2) depth, 2x iterations
Pattern quadrupled = ~1.75x (7/4) depth, 3x iterations
Pattern x8  = ~1.875x (15/8) depth, 4x iterations
Pattern x16 = ~1.9375 (31/16) depth, 5x iterations
Minibrot = ~2x depth, infinite iterations

Third order:
Pattern (forked path) = 1x depth
Pattern tripled = ~1.333x (4/3) depth
Pattern x9 = ~1.444x (13/9) depth
Pattern x27 = ~1.481x (40/27) depth
Minibrot = ~1.5x depth

Forth order:
Pattern (forked path) = 1x depth
Pattern quadrupled = ~1.25x (5/4) depth
Pattern x16 = ~1.3125x (21/16) depth
Minibrot = ~1.333x (4/3) depth

Fifth order:
Pattern (forked path) = 1x depth
Pattern x5 = ~1.2x (6/5) depth
Pattern x25 = ~1.24x (31/25) depth
Minibrot = ~1.25x (5/4) depth

Based on my observational experience, there is a clear pattern to the depth of the final minibort. For order Z=Z^N+C, the minibrot is typically at a depth of ((N+1)/N) relative to the depth of the fork point. Sometimes the final minibrot depth can be slightly less if the fork point zooms into a highly complex Julia side formation. Iteration depth is generally a more accurate metric compared to zoom depth, with each periodic double an exact multiple of the iteration depth at the fork point for order two, although I don't have hard numbers regarding iteration count for orders above two.

As a result, the higher the order, the less far you have to zoom and the less manual zooming required to locate the minibrot. Additionally, only with the order two Mandelbrot is it possible to create long rows of similar objects where the true centroid is difficult to locate visually. For an eXtreme eXample of this effect:
http://stardust4ever.deviantart.com/art/1024X-Long-595637805

Regardless, the new minibrot location method is infinitely useful for finding minitbrots in such locations rather than spending hours manually clicking or scroll-wheeling through thousands of zoom levels.

How does KF estimate the minibrot's size?

I pasted in the magic code from claude.... (i.e. I have no idea ;) )

http://ibiblio.org/e-notes/MSet/windows.htm is where I got the minibrot size estimation algorithm from.