Title: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 15, 2013, 10:01:22 PM Mandelbulber 1.18
Open source program (GNU GPL) for rendering photo-realistic 3D fractals Free download (executables and source): http://sourceforge.net/projects/mandelbulber/ Website: http://www.mandelbulber.com/ User manual: http://wiki.mandelbulber.com/index.php?title=User_Manual New features in version 1.18: - added 3D pointer in image preview window - added showing of animation path in image preview window - OpenCL - improved rendering performance of DOF - iteration loop moved out of OpenCL kernel - OpenCL - image buffer changer to float type - possible to adjust image brightness and gamma after rendering - OpenCL - implemented full Mandelbox formula - OpenCL - used CLEW for dynamic loading of OpenCL.dll (thanks Quazgaa for help) - fixed bug: lights were placed in wrong position on fish-eye mode - fixed bug: OpenCL - under Windows with nVidia gfx card the program couldn't find mandelbulber_cl_data.h file - fixed bug: OpenCL - main volumetric light wasn't rendered when no aux lights were enabled - fixed bug: OpenCL - wrong calculation of focus point for DOF - fixed bug: OpenCL - reduced noise on depth zooms - limited "delta" from very small values - fixed bug: OpenCL - program crashed on machines with low memory - added calculation of maximum queue size SVN repository: http://mandelbulber.googlecode.com/svn/trunk/ This repository is available from Google Code: http://code.google.com/p/mandelbulber/ Title: Re: Mandelbulber 1.18 - OpenCL Post by: taurus on October 16, 2013, 12:23:57 AM Great thing Budhi!
But I first finish my pending Animation on 1.17-2, cause I need the result until Friday Looking forward to testing it ASAP ;D edit> what does OpenCL - implemented full Mandelbox formula mean?My current animation looks quite full. Or do you mean that the coloring parameters are active too? Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 16, 2013, 12:46:06 AM Code: [code]:/ Windows 7 Pro x64 SP1 Nvidia GTX 260 w/ 331.40 (896MB GDDR3) 4GB ram[/code][/code] Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 16, 2013, 12:52:23 AM PS: What I'm confused by is why is it reporting GPU Memory as 0MB.
2 things: 1) I'm backing down to 327.23 on the nvidia drivers (latest 'stable') with a clean install and will try again 2) earlier if I rendered the default project without openCL then turned on OpenCL-fast and hit render again It'd work But if I change some options then try to render again the program would immediately die/quit. Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 16, 2013, 01:23:43 AM So far problem is I'm not getting expected results with OpenCL.
No OpenCL Enabled : (http://kbeezie.com/mirror/i4.jpg) OpenCL - Normal Enabled: (http://kbeezie.com/mirror/i4_ocl.jpg) OpenCL - Full Enabled : Causes the display to go black, app crashes and a popup notification stating the Nvidia driver has crashed. I'm getting basically the same with other formulas (ie: the OCL result is either just a mesh of colors as if too close or way different than the perceived scene). Thus far my results has been about 10% of the time something actually getting rendered normally, or the program crashes or shows a drastically different output. Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 16, 2013, 07:14:52 AM edit> what does mean? There are implemented also rotations for each axis. First it inculed only main rotation. So now on CPU and GPU Mandelbox look rhe same.Title: Re: Mandelbulber 1.18 - OpenCL Post by: lukesleftleg on October 16, 2013, 05:37:16 PM Hi Buddhi.
It works! And OMG it's so fast! Thank you very much indeed. :) Just a couple of small things. Firstly, I noticed that for the Tglad's Formula, the rotations are working fine, but the 'Spherical Folding Offset' doesn't seem to be switched on in OpenCL mode. There are implemented also rotations for each axis. First it inculed only main rotation. So now on CPU and GPU Mandelbox look rhe same. But from the sound of this post, it seems like you might already have it covered. Also, the Fish Eye Lens and Equirectangular Projection don't seem to be working in OpenCL either, but again, please ignore me if you know about this already. These are just small niggles though. Mostly OH WOW! Thank you Buddhi. :D Oh yeah, and it's reporting my GPU Memory as being zero as well, but it seems to work fine otherwise. (Windows 7 32 bit, Nvidia 550 ti). Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 16, 2013, 06:35:10 PM Mandelbulber 1.18 - Win64 - OpenCL now is available
download site as usual: http://sourceforge.net/projects/mandelbulber/files/? Many thanks to Quazgaa for compilation Remark: NetRender doesn't work in this version (there is some problem with WinSock under Win64) Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 16, 2013, 06:58:59 PM Was hoping it would fix OpenCL with my Nvidia GTX 260. There was bad_alloc() error. What was image resolution in this trial?Moment I click Render (typing from the 'stopped'/crashed screen as it doesn't appear in log.txt) 2) earlier if I rendered the default project without openCL then turned on OpenCL-fast and hit render again It'd work Can you run the program from command line to see what errors are in those cases?But if I change some options then try to render again the program would immediately die/quit. So far problem is I'm not getting expected results with OpenCL. OpenCL - Full Enabled : Causes the display to go black, app crashes and a popup notification stating the Nvidia driver has crashed. I'm getting basically the same with other formulas (ie: the OCL result is either just a mesh of colors as if too close or way different than the perceived scene). Thus far my results has been about 10% of the time something actually getting rendered normally, or the program crashes or shows a drastically different output. If you render default.fract (default Mandelbulb) with "Full: All Shaders" mode do get the same problem (black display, driver crash...)? Maximum zoom with OpenCL is much lower because it works on single precision floating point numbers. That's why you get "mesh of colors" when camera is too close. Not all fractal formulas are already implemented in OpenCL kernels. I suppose that's why you get different results with OCL than with CPU. Camera view can also looks different if you use "fast" and "normal" mode. It is because in those modes works only 3-point perspective. In "full" mode camera will be the same (all perspectives are implemented) Firstly, I noticed that for the Tglad's Formula, the rotations are working fine, but the 'Spherical Folding Offset' doesn't seem to be switched on in OpenCL mode. You are right. I forgot to implement it. It will be in next version.Also, the Fish Eye Lens and Equirectangular Projection don't seem to be working in OpenCL either, but again, please ignore me if you know about this already. It works in OpenCL but only in "Full: All shaders" mode Oh yeah, and it's reporting my GPU Memory as being zero as well, but it seems to work fine otherwise. It doesn't show amount of GPU memory because of some bug in OpenCL librarry (windows 32-bit version). But the application can work without this information. Only problem is that it can't adapt maximum OpenCL queue size according to avail memory. It could be a problem with older gfx card with less memory. The program could report "lack of resources". Thanks for reports! Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 17, 2013, 05:16:59 AM There was bad_alloc() error. What was image resolution in this trial? In testing never more than 960x720, some at smaller like 512x512 just to test. In fact I don't think I've tried a single OCL render higher than this yet. And I noticed that with some of the formulas that do show up, the program will crash on the 3rd time or so I render (like if I pan the camera somewhere else, then re-render). Can you run the program from command line to see what errors are in those cases? Maybe I missed it in the readme, but is there command line options specific to turning on opencl ? or is the mandelbulber-opencl when executed via CLI on windows automatically going to utilize OpenCL? (or is it settings file specific, ie: I turn it on, then save the settings). If you render default.fract (default Mandelbulb) with "Full: All Shaders" mode do get the same problem (black display, driver crash...)? Maximum zoom with OpenCL is much lower because it works on single precision floating point numbers. That's why you get "mesh of colors" when camera is too close. Not all fractal formulas are already implemented in OpenCL kernels. I suppose that's why you get different results with OCL than with CPU. Camera view can also looks different if you use "fast" and "normal" mode. It is because in those modes works only 3-point perspective. In "full" mode camera will be the same (all perspectives are implemented) ... Is there a list hiding somewhere on which formulas are provided and which are not? Also I Don't know if it's been mentioned or not, the GTX260 is one of the GPUs that supports OpenCL 1.1 (but guess they don't do 1.2 yet)? Because when I tick the checkbox for OpenCL + Full then save a setting, it doesn't seem to be reflected : Code: C:\Program Files (x86)\Mandelbulber>mandelbulber-opencl.exe -nogui -format png16 and the help output from the CLI doesn't show any specific commands to enable OpenCL. Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 17, 2013, 05:40:00 AM By the way I found a pattern that I can report with the default mandy render with OpenCL full.
Ray-trace enabled - no crash Ambient Occlusion enabled (SlowAO) - no crash Ray-trace enabled AND SlowAO enabled - black screen driver crash. (Still trying to find the command line to enable OpenCL via nogui haven't found it in forums or elsewhere). By the way in testing one of my saved files with OpenCL full, I turned off ray traced, but still got the black screen and driver crash. Attached the settings file I used. (Edit: in fact with that one in particular I could turn off all the shaders and it would still crash). Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 17, 2013, 05:50:31 AM This one also black-screen crashed on me (From one of the examples). (snapped a screenshot right before it crashed but I don't think it has anything useful there).
Some other example would be Open up Mandelbulb with Backlight, OpenCL - Full, renders normal with default settings, but if you turn off AO completely then add in a raytrace 0.5 reflection it'll black screen crash. Title: Re: Mandelbulber 1.18 - OpenCL Post by: lukesleftleg on October 18, 2013, 09:32:48 AM I've got to say Buddhi, I'm falling in love with your program all over again.
I wish I knew some C++. Maybe I should get around to learning it. Anyway, I forgot to mention, I really like the new animation path view, and the new cursor (although I'm still working out how to use the cursor). I was also going to ask for some sort of crosshair to mark the centre of the image, but I see that you've implemented one. Thanks very much for these. They're great! Re the Fish Eye/Equirectangular thing; It works in OpenCL but only in "Full: All shaders" mode Yep, found it, that's brilliant. Thank you.I was wondering though, would it be possible to have these enabled at least just in "Normal: shadows, glow, fast AO" mode too please? It's not a priority, but it's very nice to have it for planning animations. And thanks for sorting out the Spherical Offsets in the next version. I really want my next reasonable-sized project to involve these. I noticed that the "Textured background" doesn't seem to work yet in OpenCL mode. Don't worry if you know about this already, but I thought I'd better just point it out, just in case you don't. This new version is amazing though Buddhi! Thank you very much for all your hard work. As long as you don't try to go too deep, some of these OpenCL renders are good enough to use in their own animation. I thought I'd be using OpenCL mode for planning and previewing only. I'd say that the future of Mandelbulber is looking very bright indeed. Title: Re: Mandelbulber 1.18 - OpenCL Post by: lukesleftleg on October 19, 2013, 05:23:31 PM Ok, I just had to show you this.
http://www.youtube.com/watch?v=uCiT980RNOA&vq=hd720 It's nothing particularly special in terms of anything clever animation-wise, but what is special about it is the fact that it rendered overnight, and still looks pretty decent. Ok, let me say that again. That's 1800 frames, rendered at 2560x1440 (and then reduced to 720p for AA), on a not-bad-but-starting-to-show-its-age computer, and ok, it took some of yesterday evening and most of this morning too if I'm honest, but that's still quite an improvement on the nearly-a-month that it took my last animation. So that's me all happy. :) Title: Re: Mandelbulber 1.18 - OpenCL Post by: Nahee_Enterprises on October 19, 2013, 05:33:41 PM Ok, I just had to show you this. http: //www.youtube.com/watch?v=uCiT980RNOA&vq=hd720 It's nothing particularly special in terms of anything clever animation-wise, but what is special about it is the fact that it rendered overnight, and still looks pretty decent. Ok, let me say that again. That's 1800 frames, rendered at 2560x1440 (and then reduced to 720p for AA), on a not-bad-but-starting-to-show-its-age computer, and ok, it took some of yesterday evening and most of this morning too if I'm honest, but that's still quite an improvement on the nearly-a-month that it took my last animation. So that's me all happy. :) Yes, things are getting faster these days!!! Better software, better use of hardware (CPU/GPU), better machines, etc... Soon we all will be able to explore interactively the same areas that it takes us hours and days to produce for a short video. :D Title: Re: Mandelbulber 1.18 - OpenCL Post by: mclarekin on October 19, 2013, 10:32:16 PM @Mr Lukesleftleg, nothing special maybe, but still very well done. We await in anticipation what ever comes next. :)
@Mr Quaz0r, thanks for the Win 64 :) :). Minor bug is that on my computer the time elapsed/render time part, the font turns white as the render progresses, making it hard to read. @Mr Buddhi, thanks for the latest version. The new pointer is great for quickly showing the depth in the fractal. :) :) Title: Re: Mandelbulber 1.18 - OpenCL Post by: quaz0r on October 20, 2013, 01:48:57 AM @Mr Quaz0r, thanks for the Win 64 :) :). Minor bug is that on my computer the time elapsed/render time part, the font turns white as the render progresses, making it hard to read. ah sorry i hadnt actually looked at the installer for it, looks like etc/ and lib/ made its way into the win64 installer, which it shouldnt. if you have the win64 version installed to its own directory, just delete etc/ and lib/ as they are not needed and you wont have the font issue any more. etc/ and lib/ contain the theme definition and module for the win32 build so if you have your win64 build sitting in the same install directory as the win32 then you need to install it to its own directory and remove etc/ and lib/ from it. Title: Re: Mandelbulber 1.18 - OpenCL Post by: mclarekin on October 21, 2013, 08:24:59 AM Thanks Mr Quaz0r, Sir. That was the easiest "fix" ever!
Title: Re: Mandelbulber 1.18 - OpenCL, works on my linux with one minor change Post by: meander on October 21, 2013, 03:08:16 PM Thank you Buddhi! It now works on the NVidia card in my Debian Linux machine.
Fast! One minor issue. I was installing over a previous version, and the render with OpenCL was terribly random... Removing the /usr/share/mandelbulber directory and reinstalling works perfectly. Perhaps getting rid of the old stuff should be part of the installation. FAST! Thank you so much. Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 22, 2013, 09:55:49 AM Is there by any chance a way to make the app (either official or unofficial 64bit) choose my secondary graphic card for OpenCL (ie: primary is an nvidia card, but secondary is a 1GB ATI card).
When enabling OpenCL it always picks the Nvidia card. Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 22, 2013, 10:10:41 AM PS: Even with the ATI card selected as the default graphics in the bios (PEG2), the application still selects the Nvidia card when initializing OpenCL.
Edit: Would appear that after I installed the Latest Nvidia drivers, the ATI card gets selected in Mandelbulber when initializing OpenCL. (wondering if it'll stay that way). Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 22, 2013, 10:55:09 AM To update yet again (in case anyone is curious or would actually be weird enough to do the same).
I originally had my Nvidia GTX260 installed, which naturally isn't a good fit for Mandelbulber OpenCL right now. But I did have a spare (newer but 'meh' far as gaming goes) XFX Radeon R6570 (1GB DDR3). Installed (physically) the R6570, installed latest Catalyst Drivers, rebooted selected PEG2 (ATI card) as default video. Rebooted back up so both monitors were going. Then I installed the Nvidia Drivers to the latest, rebooted, selected PEG back again. Once I did that the AMD/ATI card get selected on OpenCL initialization. Even though I installed Nvidia drivers last, it would appear that OpenCL support was disabled for the Nvidia side, meaning mandelbulber will keep using the ATI card when initializing OpenCL. This is the case even if I decide to unplug the monitor from the ATI card and only use one monitor. Attached is a test render (SlowAO, Iteration Fog, Raytrace reflection) from that (which for some reason would normally crash mandelbulber if it selected Nvidia instead to render on). Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 22, 2013, 09:56:09 PM In next version I will add selection for GPUs. Can you send me (by priv message) output of clinfo command from your computer? Will be helpful for me.
Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 23, 2013, 07:24:51 AM In next version I will add selection for GPUs. Can you send me (by priv message) output of clinfo command from your computer? Will be helpful for me. Thanks I've done this, I noticed CLInfo shows both GPUs, so I guess it'll aid in coming up with a switch (would be interesting if both could be used at the same time, but I kind of doubt that with the different OpenCL implementations). Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 23, 2013, 06:07:38 PM Thanks for sending clinfo. You asked about use both GPUs at the same time. Of course it is good idea. But I will don't do this, because I have no possibilities for testing this solution. This will be some open topic for a future. Now will be only switching.
Title: Re: Mandelbulber 1.18 - OpenCL Post by: KBeezie on October 24, 2013, 05:07:43 AM Thanks for sending clinfo. You asked about use both GPUs at the same time. Of course it is good idea. But I will don't do this, because I have no possibilities for testing this solution. This will be some open topic for a future. Now will be only switching. Yea That's quite understandable, you have to be in a similar situation to be able to actively test it, but being able to switch on the fly would be very helpful, and hopefully the clinfo will help in being able to identify the various CL devices (at least the ATI implementation seems more sane especially when it'll also utilize the CPU for OpenCL where as Nvidia will not). Title: Re: Mandelbulber 1.18 - OpenCL Post by: hobold on October 24, 2013, 02:00:52 PM Thanks for sending clinfo. You asked about use both GPUs at the same time. Of course it is good idea. But I will don't do this, because I have no possibilities for testing this solution. This will be some open topic for a future. Now will be only switching. This is not a request. Not even a suggestion. Just want to mention the possibility of running OpenCL kernels on both GPU and CPU as a way to test setups with more than one compute device simultaneously active.Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on October 24, 2013, 06:44:57 PM This is not a request. Not even a suggestion. Just want to mention the possibility of running OpenCL kernels on both GPU and CPU as a way to test setups with more than one compute device simultaneously active. Thanks for this good idea. In this way I will be able to test it. Maybe I will try to do this. But now I have other priorities (e.g. possibility to use custom formulas with OpenCL). Title: Re: Mandelbulber 1.18 - OpenCL Post by: taurus on October 27, 2013, 06:34:14 PM From the beginning of the opencl development, I dreamt of using opencl versions in my half-automated searches for valuable julia sets. In Linux times, I failed to compile the packadge in a proper way, but now I think the time has come. While I could analyze about 200 parameters per night in cpu rendering mode, I can now make around 2000 parameters in acceptable quality and in full hd.
I don't know, whether this will increase quality or number of images, I produce (I doubt that), but I will surely need to find a more effective way to set up my search animations. Else I will run out of tasks soon... Thanks Buddhi for your exceptional work! ;D:beer: :beer: ;D Title: Re: Mandelbulber 1.18 - OpenCL Post by: nftaylor on November 05, 2013, 08:50:20 PM I downloaded the program dated 2013-10-16 from Sourceforge and installed it. My system is Windows7 Enterprise 64-bit SP1 with 16GB RAM. Running the program gave me an error message on startup, about a missing DLL. I tried re-installing, as advised, and got the same error message. Searching for the same DLL in my existing programs pointed to GIMP 2.8.6 <<C:\Program Files\GIMP 2\bin>>. I copied the file, ran Mandelbulber again, and got another error message, this time for a different DLL. Copy, run again. This was repeated until no error message was generated.
The list of copied DLL's: libatk-1.0-0.dll libcairo-2.dll libffi-6.dll libfontconfig-1.dll libfreetype-6.dll libgcc_s_seh-1.dll libgdk_pixbuf-2.0-0.dll libgdk-win32-2.0-0.dll libgio-2.0-0.dll libglib-2.0-0.dll libgmodule-2.0-0.dll libgobject-2.0-0.dll libgtk-win32-2.0-0.dll libharfbuzz-0.dll libintl-8.dll libjasper-1.dll libjpeg-8.dll liblzma-5.dll libpango-1.0-0.dll libpangocairo-1.0-0.dll libpangoft2-1.0-0.dll libpangowin32-1.0-0.dll libpixman-1-0.dll libpng15-15.dll libtiff-5.dll libxml2-2.dll zlib1.dll All were copied from the above noted folder, with the exception of <<libatk-1.0-0.dll>>, which was copied from <<C:\Program Files\GIMP 2\32\bin>>. Now I get to play. Thanks for your work! Title: Re: Mandelbulber 1.18 - OpenCL Post by: JohnVV on November 09, 2013, 08:58:24 AM Quote Running the program gave me an error message on startup, about a missing DLL. I am not a windows7 user ,but very longtime linux user . but win7 has a very nasty bug with programs installed to "c/:program files " it is called MS's UAC ( other software i support has MS uac issues ) it is advisable NOT to install non Microsoft programs into "c/:program files " but to C for example c:/Mandelbulber otherwise you might NEED to be the Administrator and use the "start" menu "r-click" - run as administrator- option to save config settings to the install folder in c/program files windows "should" look in the folder that has the "Mandelbulber.exe" program FIRST before it searches MS's $PATH for a dll Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on November 09, 2013, 01:28:54 PM I downloaded the program dated 2013-10-16 from Sourceforge and installed it. My system is Windows7 Enterprise 64-bit SP1 with 16GB RAM. Running the program gave me an error message on startup, about a missing DLL. I tried re-installing, as advised, and got the same error message. Searching for the same DLL in my existing programs pointed to GIMP 2.8.6 <<C:\Program Files\GIMP 2\bin>>. I copied the file, ran Mandelbulber again, and got another error message, this time for a different DLL. Copy, run again. This was repeated until no error message was generated. The list of copied DLL's: libatk-1.0-0.dll .... zlib1.dll All were copied from the above noted folder, with the exception of <<libatk-1.0-0.dll>>, which was copied from <<C:\Program Files\GIMP 2\32\bin>>. Now I get to play. Thanks for your work! Hi I tried to reproduce this error. The only way which I found is to start some older 64-bit version from C:\Program files\Mandelbulber folder. I'm not sure which version you tried to start (32-bit or 64-bit). Mu suggestion: uninstall 32-bit and 64-bit versions of Mandelbulber, clean up C:\Program files\Mandelbulber and C:\Program files (x86)\Mandelbulber folders and then install Mandelbulber from http://sourceforge.net/projects/mandelbulber/files/Mandelbulber%2064-bit%20OpenCL-1.18-Setup.exe/download I am not a windows7 user ,but very longtime linux user . but win7 has a very nasty bug with programs installed to "c/:program files " it is called MS's UAC ( other software i support has MS uac issues ) it is advisable NOT to install non Microsoft programs into "c/:program files " but to C for example c:/Mandelbulber otherwise you might NEED to be the Administrator and use the "start" menu "r-click" - run as administrator- option to save config settings to the install folder in c/program files windows "should" look in the folder that has the "Mandelbulber.exe" program FIRST before it searches MS's $PATH for a dll Mandelbulber doesn't need to write any files (config, or other) to program folder, so this problem doesn't exits with Mandelbulber. I think under Linux you also don't store any config files in /usr/bin/... This is the same. Title: Re: Mandelbulber 1.18 - OpenCL Post by: lukesleftleg on November 12, 2013, 03:40:58 PM Hi Buddhi, just a quick question.
I'm still getting a few crashes in OpenCL mode. It's not constant, and it seems to happen when I go too close to a fractal, or try vaguely controversial settings. I've just been taking the crashes in my stride, thinking that the OpenCL stuff is still in development and some instability is to be expected, but it's just occurred to me that you may want to know about these. Would it help if I provided Process Monitor reports, or do you already have enough information on the subject? Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on November 12, 2013, 06:03:34 PM Hi Buddhi, just a quick question. I'm still getting a few crashes in OpenCL mode. It's not constant, and it seems to happen when I go too close to a fractal, or try vaguely controversial settings. I've just been taking the crashes in my stride, thinking that the OpenCL stuff is still in development and some instability is to be expected, but it's just occurred to me that you may want to know about these. Would it help if I provided Process Monitor reports, or do you already have enough information on the subject? Wait for next version. Will be much more error handling and should be more resistant for errors. Then we will see what further improvement I can do. Next release should be in one or two weeks. Title: Re: Mandelbulber 1.18 - OpenCL Post by: lukesleftleg on November 12, 2013, 06:11:25 PM Wait for next version. Will be much more error handling and should be more resistant for errors. Then we will see what further improvement I can do. Next release should be in one or two weeks. Okeydoke. No worries then. :) Great news about the release schedule as well. I'm just re-rendering one of my older animations, and I reckon it'll take about that long to finish. (I've rendered most of it in OpenCL mode, but it's having some problems with the last bit, so I'm just doing those in CPU mode). Title: Re: Mandelbulber 1.18 - OpenCL Post by: magister94 on November 16, 2013, 02:06:22 AM I seem to be getting significantly different colors when using OpenCL. Is that normal? how can I make OpenCL render with the same colors?
Thanks Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on November 16, 2013, 01:30:45 PM Will be not the same. It is not optimal to use the same algorithms for OpenCL and for CPU. That's why there are some differences.
Title: Re: Mandelbulber 1.18 - OpenCL Post by: Buddhi on November 16, 2013, 09:22:05 PM Wait for next version. Will be much more error handling and should be more resistant for errors. Then we will see what further improvement I can do. Next release should be in one or two weeks. Today I have released new version. I started new thread for Mandelbulber 1.19 http://www.fractalforums.com/releases-b233/mandelbulber-1-19-opencl-user-defined-formulas-management/msg68615/#new |