Logo by Mahmut - Contribute your own Logo!
News: Support us via Flattr FLATTR Link
 
*
Welcome, Guest. Please login or register. December 16, 2017, 09:45:35 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 [2] 3 4   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: Mandelbulber v2 - 2.05  (Read 3853 times)
0 Members and 1 Guest are viewing this topic.
Buddhi
Fractal Iambus
***
Posts: 895



WWW
« Reply #15 on: October 14, 2015, 08:15:50 PM »

In mb3d formulas are executed in the selected order with the selected iteration count so you can do eg 3 iters of Mandelbox, then 1 of Menger. afterwards restarts with 3 box 1 menger etc.
Then, you have the repeat from here ... that normally is set on fo.1
So you can do 3 iters of Mandelbox, then 1 of Menger, repeat from Mandelbulb. afterwards you will calculate only mandelbulb for all following iters.
Hope it helps?
wink
So it looks like it's exactly the same. How DE is calculated for hybrids? Is it just "delta DE" algorithm?
Logged

taurus
Fractal Supremo
*****
Posts: 1175



profile.php?id=1339106810 @taurus_arts_66
WWW
« Reply #16 on: October 14, 2015, 08:31:45 PM »

interjection
- the ability to hit undo after crashing: it already saves auto-recovery information. When you restart information there should be message with question about recovering of settings. Is this doesn't work or you prefer to have something like was in od Mandelbulber?
Didn't see such a message. Maybe has to do with the fact, that I did not install v2.05, only downloaded zip file and extracted to program folder of 2.04.

Crashed reproducible after pushing an arrow button in a quick sequence (no matter what arrow). Window stays in the state like in attached jpg and remains operational (tooltips work, arrows changes coordinates etc.) but rendering does not start (0% cpu).


* xchange.jpg (143.33 KB, 1616x1038 - viewed 165 times.)
Logged

when life offers you a lemon, get yourself some salt and tequila!
Buddhi
Fractal Iambus
***
Posts: 895



WWW
« Reply #17 on: October 14, 2015, 08:32:04 PM »

as far as speed goes, when compiling for yourself you may as well add -march=native to take advantage of simd and any other instructions added to modern cpus in the last 30 years  angel  i personally also roll with -Ofast, which enables -ffast-math and the like.  i dont know how buddhi feels about that.  personally i figure if a computation is going to behave badly once in a while its going to happen anyway, id rather have the speed improvements than the strict IEEE compliance.   smiley

If anybody wants to optimize code for his machine it just enough to export CXXFLAGS=-march=native (that flags are not hardcoded in .pro file). Only which is hardcoded is -ffast-math flag (enabled by default to speed up floating point operations and doesn't make problems on different platforms).
Logged

Buddhi
Fractal Iambus
***
Posts: 895



WWW
« Reply #18 on: October 14, 2015, 08:35:15 PM »

interjectionDidn't see such a message. Maybe has to do with the fact, that I did not install v2.05, only downloaded zip file and extracted to program folder of 2.04.

Crashed reproducible after pushing an arrow button in a quick sequence (no matter what arrow). Window stays in the state like in attached jpg and remains operational (tooltips work, arrows changes coordinates etc.) but rendering does not start (0% cpu).

I see. The problem is that the program doesn't crash, but just hang. I will try to fix this problem.
Logged

zebastian
Conqueror
*******
Posts: 121



« Reply #19 on: October 14, 2015, 10:15:28 PM »

To the iteration / hybridization topic:

In the last mandelbulb3d package:
http://www.fractalforums.com/index.php?action=downloads;sa=view;down=52

under "PLEASE READ/condit tuts.zip" > unpacked > LYAP.jpg
there is a good explanation by luca on how mandelbulb3d works with custom alternation of formula based on iteration number.

since mandelbulber works with 4 separate formulas, i think it would be good to implement as:

checkbox to enable "iteration based custom formula application" (or some way better name)
editable List of elements:
each element nxm (repeat formula n m times)

so 1x7;3x3;1x2;2x3
becomes 7 iters of 1st fractal, 3 iters of 3rd fractal, 1 iter of 2nd fractal, 2 iters of 3rd fractal, repeat --->

what do you guys think?
Logged
mclarekin
Fractal Senior
******
Posts: 1739



« Reply #20 on: October 14, 2015, 11:19:34 PM »

Yes Zebastian.

My thoughts are :-

I remember that in Mandelbulber 1.## there was the options of either

repeat a sequence until end conditions are met

or

run a sequence then repeat last formula until  end conditions are met

In respect to the current MandelbulberV2 it has 4 formula tabs (V 1.## had 5)and more could be added.  More would be useful with boolean operators, however with hybrids more often means  more difficultly with quality.
Logged
DarkBeam
Global Moderator
Fractal Senior
******
Posts: 2512


Fragments of the fractal -like the tip of it


« Reply #21 on: October 15, 2015, 12:02:27 AM »

So it looks like it's exactly the same. How DE is calculated for hybrids? Is it just "delta DE" algorithm?
It is very likely! Jesse was a close follower of your program.
Internally, formulas never calculate bailout conditions. So I suspect - well I am almost sure... - they are fundamentally similar... even if he made lots of modifications during the years. And probably he even told me that he adopted deltaDe...
But I never checked that. Anyone can?
Cheers all A Beer Cup
(Jesse if you ever read this... we always miss you)
Logged

No sweat, guardian of wisdom!
zebastian
Conqueror
*******
Posts: 121



« Reply #22 on: October 15, 2015, 12:41:52 AM »

pfff, it's probably somewhere in Calc.pas
https://github.com/thargor6/mb3d/blob/master/Calc.pas

in CalcDEfull there is something about hybridization.
I am not into pascal and most certainly not into assembler shocked
Maybe buddhi can have a look...
Logged
R-TEAM
Forums Freshman
**
Posts: 12


« Reply #23 on: October 19, 2015, 09:49:49 AM »

Hi,

as i read this "..Even running OpenCL on my old  cpu was faster, but then it is only single precision." i think, it is not possible to get OpenCL with double precision ??

Regards
R-TEAM
Logged
mclarekin
Fractal Senior
******
Posts: 1739



« Reply #24 on: October 19, 2015, 11:21:46 AM »

Mandelbulber Version 1.21 has OpenCL single precision.  I think it may have been because of the limited amount of GPUs that could run double precision .

 I do not know if when OpenCl is added to MandelbulberV2, whether it will be double precision, but I hope there is a way to choose between single or double precision. For me the speed of single precision OpenCL is what I like, I assume double precision would be a bit slower.
Logged
quaz0r
Fractal Molossus
**
Posts: 652



« Reply #25 on: October 19, 2015, 03:05:03 PM »

s/bit/lot/
Logged
mclarekin
Fractal Senior
******
Posts: 1739



« Reply #26 on: October 20, 2015, 12:29:00 AM »

@quaz0r grin

Complex not that seems smiley smiley smiley
Logged
R-TEAM
Forums Freshman
**
Posts: 12


« Reply #27 on: October 20, 2015, 12:53:34 PM »

Hi,

it depends on the GPU ...
I use pro 3D GFX with Maya and other high end programms - so i have an (new) FirePro w8100 .. enough double precision power smiley

Regards
R-TEAM
Logged
zebastian
Conqueror
*******
Posts: 121



« Reply #28 on: October 20, 2015, 01:22:34 PM »

it would have been nice if double precision on gpus behaves like:
performance(dp) = (1 / 2) * performance(fp)
but this is only the case with certain high end AMD cards. Lucky you, R-TEAM  afro
with consumer cards its more like 1/3 - 1/8 or even far worse with Nvidia cards.

here is a good summary on this topic
http://arrayfire.com/explaining-fp64-performance-on-gpus/

when it gets to implementing new opencl shaders we will certainly take double precision into consideration.
Logged
R-TEAM
Forums Freshman
**
Posts: 12


« Reply #29 on: October 20, 2015, 01:42:33 PM »

Hi,

mmhh .. lucky .... it depends on the definition of lucky ...
The card is VERY hugh .... it is not quiet if i need power (thinking of an water cooling system ...) and i get on the most apps
only an marginal benefits over an consumer card - in DX it is slower as the actual cosnumer top modell - and even this is much cheaper ..
In openGL it depend on the app - maya is at best 20 times as fast as with an consumer card .. this is nice - but only few apps get this hugh performance boost from the highly optimized drivers .. most times i get 1,5-2 times faster as the "same" consumer GPU.
On the good side, it is nice to have GFX on an 4K 60Hz display in 10bit ... and i get an good support from AMD if i need it..
But this to an very expensive price tag .. so i am unsure if i lucky smiley  [but i need an pro card for pro gfx editing/3D constructing, so it
was more an, "i musst buy an pro card - and then, it should have POWER smiley " thing ... ]

i would have more on the positive side if i can use the GPU dp power .... (i hope in the future the rendering in maya or mental ray or renderman get an boost with GPU use ..this was one of the points, the card make sense, even with this price ..)

Regards
R-TEAM
Logged
Pages: 1 [2] 3 4   Go Down
  Print  
 
Jump to:  

Related Topics
Subject Started by Replies Views Last post
Mandelbulber 0.85 Mandelbulber Buddhi 6 2130 Last post July 25, 2010, 10:00:13 PM
by kram1032
Mandelbulber 0.93 Mandelbulber Buddhi 12 3598 Last post October 17, 2010, 03:19:15 PM
by Buddhi
mandelbulber Help & Support ramblerette 1 453 Last post October 18, 2010, 02:56:02 PM
by ramblerette
Mandelbulber 0.94 Mandelbulber « 1 2 » Buddhi 15 4038 Last post October 24, 2010, 09:36:01 AM
by Buddhi
Mandelbulber 1.12 Releases Buddhi 13 3816 Last post June 23, 2012, 08:44:32 PM
by visual.bermarte

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 0.228 seconds with 29 queries. (Pretty URLs adds 0.017s, 2q)