Mandel Machine

[Botond Kósa]

you know guys, if this "glitch" works and doesn't do any harm, PLEASE don't see this as a glitch or a mistake you need to correct!
See it as Evolution! There has been a strange mutation in the formula! creating something new!

You should feature it! Try to push it further! Try to make it repeat and evolve itself!
A really NEW branch of the M-Set. And you want to "correct it"?! Noooooooooooooo!!!!

Why always "perfect" things.
zooming into the m-set is so diverse, but in can become boring, as the main patterns always nearly stay the same

THIS is something DIFFERENT.
I see it as EVOLUTION!
By random mutation, just as it happens in the real world.
Mistakes aren't bad, if you make the best out of them and improve on them.
Now Botond, please, make the evolution conscious and don't kill of the first spark! Let it grow!

Interesting idea, but I am a bit skeptical about the reproducibility of these mutations. When I first discovered them I found that altering the location of the reference pixel in such a mutated image by just 1 pixel can lead to completely different mutations. But maybe it could be possible zoom into such a mutation by reusing the reference from the last image. I'll give it a try.

[Chillheimer]

yeah!!!

I found that altering the location of the reference pixel in such a mutated image by just 1 pixel can lead to completely different mutations.

I hear a Butterfly scream.. what would Lorenz say to this?

Worlds within worlds. within one pixel...

[Botond Kósa]

A new beta version, 1.2.5 is available. I made the previous betas downloadable in case the latest has major issues. I am going on a final summer holiday tomorrow and won't have net access for the next week.

[Dinkydau]

6000 zooms, time for another record deepest julia morphing?

[Dinkydau]

The glitches that stardust4ever posted also occur in the latest version, for example on the attached location. Using fewer terms of series approximation may help. Disabling series approximation always works, but that's much, much slower...

[stardust4ever]

The glitches that stardust4ever posted also occur in the latest version, for example on the attached location. Using fewer terms of series approximation may help. Disabling series approximation always works, but that's much, much slower...

I have solved the glitch problem. Much like the base Mandelbrot formula, at ever increasing zoom depths, it eventually becomes necessary to increase the precision not only on the base pixel but also increase the sliding scale precision of the perturbation approximation series. Fortunately Mandel Machine also allows the manual selection of the scale bits precision. Switch this from automatic to manual and select a higher scale precision from default and the anomaly clears right up. Instead of pixelation at the fringe bit depth like with real precision, you get perturbation errors. They kind of remind me of the look of some rendered Mandelbrot images created by manually Manipulating the seed C0 to some value other than C or 0.

I have another glitch I'd like to report that seems to affect version 1.2.4 (I've not downloaded 1.2.5 yet). The saved PNG files seem to be off spec in some way. GIMP reports the output PNGs as corrupt, however MS Paint and Image Viewer render the PNGs correctly. Workaround is to open the image in MS Paint, save, then reopen in GIMP or other image editor for further editing. I typically use Bicubic to scale photos but bilinear option seems to work better for fractal data provided the image is scaled at integer ratios. For instance you can render a huge image at 12800x9600, save as PNG, then scale down to 3200x2400 Bilinear for a nice 4x4 antialiased image at decent print resolution, but still be able to view the finer fractal details in the source image. This nets the same final results as rendering 3200x2400 at 4x4 AA and takes the same length of time.

Finally, for those of you who want to create your own color palettes, I have discovered two backdoor methods for inserting palette data into an image. If Mandel Machine does not yet have an option for tweaking color palettes, maybe a menu option for importing Kalles Fraktaler palette files could be added. Read on for the backdoor method of importing palettes into Mandel Machine.....

Method #1: Use Kalles Fraktaler to create custom color palette by setting your preferred colors within the colors editor. Load arbitrary coordinate data into the image. The location does not matter but for some reason you need to zoom in at least a tiny bit for Mandel Machine to properly load the file. Save. Open the Kalles Fraktaler parameter file in Mandel Machine. Use the save dialog and uncheck everything but the Palette checkbox. Save the file as [name of color scheme].mmf Next, load your favorite mmf parameter into Mandel Machine. Next, load [name of color scheme].mmf over the existing parameters. Make sure only the Palette checkbox is highlighted. If you saved only the paletted data, the other options should be grayed out. Upon load, all your current parameter settings will remain intact, but your favorite custom Kalles Fraktaler color scheme will be loaded. Enjoy rendering your image with a custom color palette!

Method #2: Open the Mandel Machine using a plain text editor and manually edit the color values. For pros only.

[SeryZone]

Botond, when you end work for normal glitch-correction? Unable to render smth! Please, turn it on!

[Botond Kósa]

Tomorrow!   

[Botond Kósa]

Summer's gone, time for a new release: v1.2.6 brings the long-awaited glitch correction based on Pauldelbrot's algorithm. It is still disabled when working with 80-bit extended precision, but these occasions should be rare thanks to series approximation enabling the use of data types with lower scale. In order to avoid calculating too many reference points, when the remaining glitches contain no interior points, their pixels are interpolated from the neighboring non-glitch pixels.

The problem of distorted Julia centers is still not fixed completely, so I introduced a series approximation limiter: a maximum number of skipped iterations can be set in the Computation box.

Thanks to stardust4ever for pointing out a bug in the png saving routines. The CRC32 checksums were miscalculated, and some image editors that verify checksums could not open the PNGs.

For a full list of changes, see http://web.t-online.hu/kbotond/mandelmachine/#changelog

[Kalles Fraktaler]

Great news, I hope you will succeed in completing the fastest and correct Mandelbrot explorer in the universe!

However I cannot start 1.2.6 - nothing happens.
MandelMachine.exe *32 flashes in Windows Task Manager processes and then disappear immediately.
I was just about to ask you if you had accidentally put a 32-bit executable in the zip.
But when I reverted to 1.2.5 again and it starts without problem, there is no MandelMachine.exe in Task Manager, the main window points to the process javaw.exe. So I guess MandelMachine.exe is just some kind of startup program?

Anyway, unfortunately I cannot give any better details than that.
On my laptop I am running Windows 2008 server.

[Botond Kósa]

However I cannot start 1.2.6 - nothing happens.
MandelMachine.exe *32 flashes in Windows Task Manager processes and then disappear immediately.
I was just about to ask you if you had accidentally put a 32-bit executable in the zip.
But when I reverted to 1.2.5 again and it starts without problem, there is no MandelMachine.exe in Task Manager, the main window points to the process javaw.exe. So I guess MandelMachine.exe is just some kind of startup program?

Yes, MandelMachine.exe is just a wrapper around the .jar file that contains the application. When started, it searches for a suitable Java installation (version 1.7.0 or newer, 64-bit) and launches the javaw.exe, and then terminates. What you see with 1.2.5 is exactly the expected behavior.

Assuming you made no changes to the Java configuration between trying the different MM versions, my guess would be that something is going wrong again in the native part of the code (mm64.dll), causing the whole JVM to crash before the application window appears. Did it generate an error report in a file named hs_err_pid*.log? I have currently no access to my non-AVX computer, and maybe some AVX instructions were accidentally inserted in the SSE codepath.¹

I'm going to release a debug version with detailed logging to help diagnose these problems.

edit:
¹ I made the assumption that your CPU is not AVX-capable from one of your earlier comments. If it is, then something else is causing the crash.

[Kalles Fraktaler]

No error file is created at least in the same directories as the executables...

[Botond Kósa]

I have uploaded a new, restructured 7z package for version 1.2.6. It allows more control over how the application is launched. The jar containing the Java classes is extracted from the MandelMachine.exe file and placed in a separate mm.jar. The exe file is just a launcher now. A cmd file is also included that launches java.exe with mm.jar and displays the log in a console window. The cmd can be edited to redirect the standard output to a file (by appending '> filename' to the end) or to specify an exact path to java.exe.

[Botond Kósa]

I found the cause for the crash in v1.2.6: since the last published version I switched from Visual C++ 2010 to 2013, so the msvcr100.dll runtime had to be replaced by msvcr120.dll in the installer. With Visual Studio installed on my computer, I could run the program without the separate dll, so this dependecy remained unnoticed.

I uploaded the new 7z package, hope it works for everyone.

[stardust4ever]

I'm still having issues with the application crashing when using the scroll wheel on the mouse to zoom in and out. I believe what is happening is when rolling the scroll wheel in or out multiple clicks, the image starts to render, then the render process is interrupted by successive wheel increments, causing the application to crash. I'm glad you disabled adjusting the settings during render in version 1.2.6; this will prevent a lot of hangs.

However, I'm almost scared to use the scroll wheel to zoom in and out now, because I'm afraid I will crash the app and lose my progress. I think I crashed the app three times in a couple minutes trying to zoom in/out with the scroll wheel. Scrolling is far more intuitive and easier than creating a bounding box with the mouse and clicking to zoom in and out so I hope you get this fixed.