Kalles Fraktaler 2 + GMP

  • 232 Replies
  • 11669 Views

0 Members and 1 Guest are viewing this topic.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #195 on: February 27, 2018, 11:12:55 PM »
No, I assumed you attached the crashing case, not something that I have to edit... sorry for misunderstanding...

I edited it and still no crash.  Maybe a phantom memory corruption issue (a bug in kf) that is system dependent...

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #196 on: February 28, 2018, 12:19:01 AM »
I edited it and still no crash.  Maybe a phantom memory corruption issue (a bug in kf) that is system dependent...
It consistently crashes on my windows 7 desktop, but runs fine on my windows 10 laptop. Oh well...

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #197 on: March 02, 2018, 04:16:52 AM »
Could we get uniform "stay within pixel jitter" for the next release?
See for example https://computergraphics.stackexchange.com/questions/4248/how-is-anti-aliasing-implemented-in-ray-tracing.
Approximate "Blue noise" which seems optimal can then be gotten by oversampling the jittered images.
Here's another reference: https://graphics.stanford.edu/courses/cs248-99/samp/samp2.html
« Last Edit: March 02, 2018, 04:39:22 AM by gerrit »

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #198 on: March 03, 2018, 01:01:34 AM »
Problem with vertical lines appearing when zooming past around E2000 at various locations, sample attached. Fine in 2.12.9.1
Goes away when selecting "use floatexp always".

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #199 on: March 03, 2018, 07:36:31 PM »
Problem with vertical lines appearing when zooming past around E2000 at various locations, sample attached. Fine in 2.12.9.1
Goes away when selecting "use floatexp always".
Thanks for the location.  Can reproduce.  Not related to DE colouring.  The iteration counts in the lines are very low (1 more than the series approximation iteration). Seems to go away when enabling jitter. No real idea why this is happening yet...

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #200 on: March 03, 2018, 08:55:00 PM »
Nitpick: when saving kfs extension in file save dialog  is kfr and you have to remember to change it.
Also, a bit more nasty, if you then later "save" (Ctr-S) the kfs file is overwritten with what should go to the kfr file, so remember to do "save as" and change extension back after you saved a setting.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #201 on: March 09, 2018, 06:46:27 AM »
Also, a bit more nasty, if you then later "save" (Ctr-S) the kfs file is overwritten with what should go to the kfr file, so remember to do "save as" and change extension back after you saved a setting.

I think I fixed this and some other things that could have changed where Ctrl-S would save to (the original code was a bit of a mess using fixed size char arrays, I changed it to use std::string some time ago but didn't clean up the "everything shares one global buffer" aspect - now there are local buffers for most things, and two global buffers, one each for location and settings).

Couldn't reproduce the extension issue, maybe I fixed in the process, maybe not.  In WINE the file save dialog doesn't have any extension prefilled.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #202 on: March 09, 2018, 03:00:00 PM »
Problem with vertical lines appearing when zooming past around E2000 at various locations, sample attached. Fine in 2.12.9.1
Goes away when selecting "use floatexp always".
Seems maybe to be something to do with delta x or y being nearly exactly aligned with a reference.  Possibly some bad behaviour (maybe something with floatexp zero is broken? don't know yet).  I got some horizontal lines at 1e1000 so it isn't super-deep specific or vertical specific.  Recommended workaround is to use jitter, seems it helps as much as using floatexp in my tests...

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #203 on: March 12, 2018, 07:38:20 PM »
maybe something with floatexp zero is broken?
Yes, something was broken indeed - very small "in64_t" exponents would overflow "int" in "ldexp()" calls, I added checks for these cases and now no more lines.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #204 on: March 12, 2018, 07:45:45 PM »
kf-2.12.11 released!  get it from https://mathr.co.uk/kf/kf.html

- uniform jitter by default (suggested by gerrit)
- Gaussian jitter available as an option
- jitter scale factor setting (suggested by gerrit)
- log verbosity flag for command line rendering
- upgrade to libjpeg6b2 from http://jpegclub.org/support/
- fix bug where Ctrl-S would overwrite the wrong file (reported by gerrit)
- fix bug where lines would appear aligned with references (reported by gerrit and CFJH)
- fix "color method and infinite waves are immediately applied" even when colouring is disabled (reported by gerrit)

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #205 on: March 30, 2018, 02:32:07 PM »
kf-2.12.12 released! get it from https://mathr.co.uk/kf/kf.html

* Free Software under GNU Affero General Public License, version 3 or greater
* least squares numerical differences (experimental; subject to change)
* Laplacian numerical differences (suggested by gerrit)
* (possibly stratified) tiled settings/parameters generator program
* stratified tiling reassembler GNU Octave script

kf-2.13.1 released! see https://fractalforums.org/kalles-fraktaler/15/kalles-fraktaler-2-13/1123/

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2161
« Reply #206 on: March 30, 2018, 08:10:19 PM »
Thanks very much for all that. Looking forward to tryng it out!

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #207 on: May 15, 2018, 09:27:23 AM »
With offline rendering if I change the SA terms to "not auto" and "terms" 30 I get an immediate segmentation fault. Example attached. Lowering resolution to 640 and it doesn't crash.

I have a reproducing test case now, it seems to crash after the reference iterations are computed before the series approximation is computed, only when not-auto (terms=60 in my case).  Possibly upgrading WINE exposed the bug in KF.  Looks like a buffer overrun / memory corruption error, the m_nPixels[] pointers are set to garbage... even though they were initialized correctly.  Debugging...

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #208 on: May 15, 2018, 10:47:20 AM »
kf-2.12.13 released!  get it from https://mathr.co.uk/kf/kf.html#download

just one small fix, for the crash with manual series approx terms in command line rendering mode

fix is also in the 2.13 git branch, not released yet

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1540
    • mathr.co.uk
« Reply #209 on: June 01, 2018, 01:02:36 PM »
kf-2.12.13 released!  get it from https://mathr.co.uk/kf/kf.html#download

This version has a severe performance bug.  I have the fix but no time to release until next week...


xx
Kalles Fraktaler 2.13

Started by claude on Kalles Fraktaler

90 Replies
4661 Views
Last post March 20, 2019, 05:32:26 AM
by claude
xx
Kalles Fraktaler 2.14

Started by claude on Kalles Fraktaler

233 Replies
7386 Views
Last post July 07, 2020, 05:59:03 AM
by gerrit
xx
Another version of Kalles Fraktaler?

Started by greentexas on Kalles Fraktaler

9 Replies
1099 Views
Last post September 22, 2017, 02:59:16 PM
by greentexas
xx
Feature ideas for Kalles Fraktaler

Started by FractalAlex on Kalles Fraktaler

11 Replies
179 Views
Last post June 05, 2020, 04:01:18 PM
by FractalAlex
xx
Kalles Fraktaler Movie Maker64

Started by Bill Snowzell on Fractal movie gallery

0 Replies
297 Views
Last post February 01, 2018, 09:29:04 PM
by Bill Snowzell