Logo by Sockratease - Contribute your own Logo!

END OF AN ERA, FRACTALFORUMS.COM IS CONTINUED ON FRACTALFORUMS.ORG

it was a great time but no longer maintainable by c.Kleinhuis contact him for any data retrieval,
thanks and see you perhaps in 10 years again

this forum will stay online for reference
News: Visit us on facebook
 
*
Welcome, Guest. Please login or register. April 16, 2024, 06:45:21 PM


Login with username, password and session length


The All New FractalForums is now in Public Beta Testing! Visit FractalForums.org and check it out!


Pages: [1]   Go Down
  Print  
Share this topic on DiggShare this topic on FacebookShare this topic on GoogleShare this topic on RedditShare this topic on StumbleUponShare this topic on Twitter
Author Topic: Filamentous  (Read 733 times)
0 Members and 1 Guest are viewing this topic.
Pauldelbrot
Fractal Senior
******
Posts: 2592



pderbyshire2
« on: March 08, 2014, 05:12:23 AM »

Filamentous



http://www.fractalforums.com/index.php?action=gallery;sa=view;id=15698

A Mandelbrot deepzoom calculated with Nanoscope.

This image is close to a very tiny minibrot (4x10-289 or so wide) and has magnification 10-291.

Nanoscope's optimizations for working mostly with just ordinary doubles without extended exponent precision had caused problems zooming a few more times into the center of this image, but no longer. There is now a bailoutlike test in the inner loop to check for the doubles to become "denormalized". If necessary, the iteration is redone and the next performed with a rescaled exponent, before returning to the previous scale, to get over a "hump" caused by a reference orbit point of very small magnitude. If the reference orbit point is so small as to itself denormalize or even snap to zero when converted to a double, the iteration is instead done using 16-bit arbitrary precision, which is slower but has nigh-unlimited exponent range, using the arbitrary precision version of that reference orbit point.

These changes appear to have fixed the deeper images below this zoom that I was using as test cases for debugging this behavior. Most iterations are still done using just a small number of doubles with normal exponent range.

Unfortunately, there are still some things to tweak in the automatic glitch correction, so Nanoscope is still not ready for prime time. But progress is still being made.
Logged

alij
Conqueror
*******
Posts: 102


« Reply #1 on: March 08, 2014, 11:14:24 AM »

Lovely image.

I like the way the pink shapes jump out of each other.
Logged
Dinkydau
Fractal Senior
******
Posts: 1616



WWW
« Reply #2 on: March 08, 2014, 01:49:27 PM »

Yes, I like that too.
Logged

Pauldelbrot
Fractal Senior
******
Posts: 2592



pderbyshire2
« Reply #3 on: March 09, 2014, 08:58:43 AM »

Thanks!
Logged

Kalles Fraktaler
Fractal Senior
******
Posts: 1458



kallesfraktaler
WWW
« Reply #4 on: March 09, 2014, 11:39:34 AM »

Nice image!

Ordinary double should be fine up to e300, but if you have a large bailout value it might be lower. What bailout do you use?
Note that I am using positive exponent when mentioning magnifications.
Logged

Want to create DEEP Mandelbrot fractals 100 times faster than the commercial programs, for FREE? One hour or one minute? Three months or one day? Try Kalles Fraktaler http://www.chillheimer.de/kallesfraktaler
http://www.facebook.com/kallesfraktaler
Pauldelbrot
Fractal Senior
******
Posts: 2592



pderbyshire2
« Reply #5 on: March 10, 2014, 04:38:53 PM »

Nice image!

Ordinary double should be fine up to e300, but if you have a large bailout value it might be lower. What bailout do you use?
Note that I am using positive exponent when mentioning magnifications.

100000.

But I have it working down to e768 and beyond. Had some bugs to fix along the way but I confirmed two nights ago that it's seeing the same shapes as UF slowly and painstakingly starts to show at the same spot and depth. I just work with a renormalized delta value and track the exponent elsewhere. Usually it can go around 500 iterations before needing to rescale everything, though it has to right away if the new values denormalize or go to zero (especially if the double version of the reference orbit point is zero). Then one iteration is (re)calculated with a 16 digit java.math.BigDecimal and a rescaling done straight after. If I disable this handling, it "derezzes" at around e324, and if I have the handling but have a certain recently-swatted bug in the renormalization, it develops an X-shaped artifact at e748 and is blank by e750. With all this fixed, the image at e768 is confirmedly correct, and it's definitely exercising all of these cases, including one iteration around 50,000 having the reference orbit point smaller than e-312 but bigger than zero, which makes the double version snap to zero; the points in the e768 image all take over 110,000 iterations. The iteration with the very tiny reference orbit point is because the minibrot just below and left of the image that started this thread is as small as it is; that iteration is equal to the minibrot's period.
Logged

rahulmukerji
Safarist
******
Posts: 81


« Reply #6 on: March 15, 2014, 05:58:05 AM »

Stunning colors on this piece !
Logged
Pauldelbrot
Fractal Senior
******
Posts: 2592



pderbyshire2
« Reply #7 on: March 16, 2014, 04:01:04 AM »

Stunning colors on this piece !

Thanks! Product of my multiwave system of gradient generation, of course.
Logged

Pages: [1]   Go Down
  Print  
 
Jump to:  


Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Valid XHTML 1.0! Valid CSS! Dilber MC Theme by HarzeM
Page created in 0.175 seconds with 23 queries. (Pretty URLs adds 0.009s, 2q)