• March 02, 2021, 05:36:12 PM

Login with username, password and session length

Author Topic:  Kalles Fraktaler 2.14  (Read 13671 times)

0 Members and 1 Guest are viewing this topic.

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #285 on: November 11, 2020, 12:32:36 AM »
kf-2.14.10.4 released
get it at https://mathr.co.uk/kf/kf.html#kf-2.14.10.4

one bug fix in the 2.14 stable branch: the "solve glitch by random choice" wasn't random enough when used in command line rendering mode

the rest of the changes are build-system related (switch mingw threading model from win32 to posix; and the 32bit build is back)

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #286 on: November 19, 2020, 12:03:21 AM »
kf-2.14.10.5 released
get it at https://mathr.co.uk/kf/kf.html#kf-2.14.10.5

stable branch updates: the 32bit build no longer crashes on startup with an assertion failure, and scaled long double rendering is fixed.  also upgraded gmp to 6.2.1.

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #287 on: November 24, 2020, 07:52:15 PM »
kf-2.14.10.6 released
get it at https://mathr.co.uk/kf/kf.html#kf-2.14.10.6

stable branch updates:

    - fix long double rendering for formulas with abs (reported by Mr Rebooted)
    - fix floatexp rendering for fractional bailout norm power
    - fix some float conversion warnings and potential problems
    - make prepare script modular (second argument determines libraries to build, default is all libraries)
    - patch OpenEXR to try to fix issues with non-ASCII paths on Windows (reported by Ookamitada Shirohi)

donations to support my work are welcome via https://mathr.co.uk/kf/kf.html#donate

Offline saka

  • Fractal Phenom
  • ****
  • Posts: 47
Re: Kalles Fraktaler 2.14
« Reply #288 on: December 04, 2020, 11:17:10 PM »
Hi Claude,
Something I've just noticed, but been there a long time... 
Why are positive imaginary values below the real-axis, and negative values above?  (Location dialog, and .kfr files)
Non-symmetrical fractals like the Burning Ship look fine though.
Cheers

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #289 on: December 05, 2020, 12:21:00 PM »
Why are positive imaginary values below the real-axis, and negative values above?
Historical accident.  Changing it now after all these years would be confusing and annoying.  Adding an option to flip would be a possibility I suppose...

Offline FractalAlex

  • Fractal Frogurt
  • ******
  • Posts: 487
  • Experienced root-finding method expert
Re: Kalles Fraktaler 2.14
« Reply #290 on: December 05, 2020, 12:27:37 PM »
Historical accident.  Changing it now after all these years would be confusing and annoying.  Adding an option to flip would be a possibility I suppose...
At one point, I even had the same thought. Flipping horizontally or vertically.
"I am lightning, the rain transformed."
- Raiden, Metal Gear Solid 4: Guns of the Patriots

Offline Dinkydau

  • Fractal Furball
  • ***
  • Posts: 280
    • DeviantART gallery
Re: Kalles Fraktaler 2.14
« Reply #291 on: December 07, 2020, 10:55:43 PM »
I zoomed to a location with NR-zooming and then saved a kfr file. When I open the file the depth in the location window is E308 instead of E5877 what it should be. Rendering fails. I get a black image. An easy way to reproduce the problem: start the program (2.15.1.5 or 2.14.10.6). Set the depth in the location window to 1.0000000000000000E5800. Zoom out with the scroll wheen. The depth will turn into E308 or 0.0E0 which is not correct.

Is the speed of the 2.14 version still better than 2.15? I consider using kf-2.14.7.1 for my render because that's the last version before scaled long double became broken and it doesn't have the bug described above. Any idea what's the most stable version for scaled long double rendering that also has the most important speed improvements?

I just want to render Mandelbrot power 2 at E5877; no special features required.

Offline saka

  • Fractal Phenom
  • ****
  • Posts: 47
Re: Kalles Fraktaler 2.14
« Reply #292 on: December 07, 2020, 11:13:47 PM »
Adding an option to flip would be a possibility I suppose...

haha,  I figured it was historical. Curiously it seems to be compatible with some old Xoas coordinates I found on the net, so maybe that is why it was done.

It's kind of confusing when sharing co-ordinates with other non-KF users through. 

May I suggest that you flip it when moving to the 2.16 version of KF. But change the file format slightly so that you can always maintain backwards compatibility for .kfr's.

For example, instead of:
Re: 0.443269781
Im: 0.3752443

Use something like
Real: 0.443269781
Imaginary: -0.3752443

For version 2.16 itself, you could write both versions, so that you have forwards compatibility for a while also.

Just some thoughts,  I'm not too worried. 

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #293 on: December 07, 2020, 11:23:59 PM »
Thanks Dinkydau for the bug report, I will see if I have time to fix it this week.  I can reproduce it: set location to -2 + 0 i @ 1e5087, increase bailout iterations to see image, scroll wheel out -> status bar displays i.nfe002 and blank image (all white in my case), open location window gives zoom 1.7976931348623277E308.  My guess is something is being mis-converted via double somewhere which is losing range (overflowing to inf or underflowing to 0, depending which way you look at it).  EDIT the brokenness (debugging printouts hint at multithreading-related memory corruption (maybe?)) seems to have occurred between kf-2.14.10.3 and kf-2.14.10.4, and the only relevant change is rebuilding with POSIX threading model for MinGW instead of WIN32.  I will try reverting that. EDIT2 looking at Microsoft documentation, I see
Quote
A thread in an executable that calls the C run-time library (CRT) should use the _beginthreadex and _endthreadex functions for thread management rather than CreateThread and ExitThread
which may be causing further trouble...  TODO redo all the threading to use C++ std::thread etc which should hopefully do the right thing...

Was the depth saved correctly in the KFR (check with Notepad or other text editor)?

There should be little or no speed difference between 2.14.latest and 2.15.latest (unless using OpenCL new in 2.15, which may be faster/slower depending on lots of things, but for zoom depths of scaled long double it is probably much slower than non-OpenCL).  If anything, 2.15.latest might be a fraction faster for scaled (long) double, because a couple of numerical operations were optimized in the inner loop.

The last big speed problem was between 2.12.something and 2.14.something (i.e, 2.13 was slower, because of an extra (constant, but not visibly so to the compiler) branch in the inner loop that was not floated outwards: "if for else for" is much faster than "for if else" (at the cost of larger code size).
« Last Edit: December 08, 2020, 02:16:17 AM by claude, Reason: update »

Offline claude

  • 3f
  • ******
  • Posts: 1787
    • mathr.co.uk
Re: Kalles Fraktaler 2.14
« Reply #294 on: December 08, 2020, 03:00:24 AM »
kf-2.14.10.7 released
get it at https://mathr.co.uk/kf/kf.html#kf-2.14.10.7

stable branch updates:

    - revert 2.14.10.4's threading model change: back to win32 from posix (fixes weirdness like zooming out resetting zoom to infinity or 0) (reported by Dinkydau)

donations to support my work are welcome via https://mathr.co.uk/kf/kf.html#donate

Offline Dinkydau

  • Fractal Furball
  • ***
  • Posts: 280
    • DeviantART gallery
Re: Kalles Fraktaler 2.14
« Reply #295 on: December 08, 2020, 05:12:56 PM »
Thanks Claude. I'll try 2.15.1.6.

Was the depth saved correctly in the KFR (check with Notepad or other text editor)?
Yes, it's correct in the file.


xx
Kalles Fraktaler 2 + GMP

Started by claude on Kalles Fraktaler

232 Replies
13429 Views
Last post September 12, 2019, 10:27:53 PM
by claude
xx
Kalles Fraktaler 2.13

Started by claude on Kalles Fraktaler

90 Replies
5179 Views
Last post March 20, 2019, 05:32:26 AM
by claude
xx
Kalles Fraktaler 2.15

Started by claude on Kalles Fraktaler

115 Replies
4707 Views
Last post Yesterday at 07:51:00 PM
by FractalAlex
xx
Another version of Kalles Fraktaler?

Started by greentexas on Kalles Fraktaler

9 Replies
1306 Views
Last post September 22, 2017, 02:59:16 PM
by greentexas
xx
The history of Kalles Fraktaler - comparison between old and new

Started by FractalAlex on Kalles Fraktaler

10 Replies
360 Views
Last post June 11, 2020, 03:19:33 AM
by FractalAlex