Welcome to Fractal Forums

Hi !

This software need some serious OpenCL/CUDA turbo boost. is it planned ? (or should i buy a 48cores monster ?  :dink: )

Not planned, sorry!

It is not that easy to change everything... maybe easier to make a new proggy (David Makin has something in mind i guess).

Definitely, and your program works in my granny pc so don't change it :embarrass: :dink:

Am working on it....along with several other projects.
At the moment (apart from "real" work for Parys Technografx) I'm concentrating on getting a real-time landscape engine up and running using shader 2 code (i.e. compatible with iOS).
I'm just working on an optimum shader 2 version of the diamond-square algorithm that requires no expensive run-time randoms, hopefully zero branches and source data of just 5 values per bit of horizontal coordinates. Hopefully even on iOS 4 it will be fast enough to use the algorithm multiple times so that it not only gives an fBm landscape for the heights but also for things like ground-type, vegetation etc.
The rendering will be done in shader 2 using an adaptation of distance estimation (as myself, Ron and Jos have done for escape-time fractal heightfields).
Of course it may turn out too slow on iOS (given that a typical shader 2 app that runs at unrestricted 200+fps on a full-video card runs at say 15-20fps under the iOS video).
As for the shader 2 fractal software (UF-style but no deep zooms due to float restriction) that will just run as fast as it can and if not "real-time" under iOS it will at least be useable time, in fact comparable to say UF itself on a quadcore I'd guess.

For Parystec I'm just converting "i.Candy" from iOS to OSX, you'd think it'd be easy but I ran into problems putting normal views over an OpenGL View under Snow Leopard - in this regard even iOS 2 was more advanced. Apparently the issue is fixed in Lion but that doesn't really help when for quite a while most users will be on Leopard and Snow Leopard. To those who know that I've been working on this a while my excuses are that it's my first ever OSX program and at the moment I'm suffering from partial "burn-out" (think "writer's block") :(

I'm awaiting a dual 6-core I7 MacPro, so that's 24 including the hyperthreading ;)
Due by 4th May.
Now I have 3 old (and I do mean oooold) Windows PCs to convert into a single one to make room !
( e.g. one of them is a P2 266MHz with 2 hard drives each less than 10GB and a (non-3D) Matrox video card - the only good thing is the Creative Labs AWE64 card but of course that is ISA).

David, sounds really like a lot of work, make sure to get your head free from time to time.
I don't program on more than one project for a longer time, but i am not a real programmer, so i havn't.

Of course, some GPU renderers already exists, and so far i know is scripting included in some shader languages.
No need to program all by yourself maybe...

Well I'm basically doing it "the hard way" to get my head round shaders generally before progressing on to full OpenCL etc.
Also of course at the moment full OpenCL is not available for iOS.
I never did any sophisticated OpenGL or DirectX programming before doing the free version of i.Candy for the iPhone - in fact the most I'd done using GPU's was some basic texture blitting in our PT Pocket Office VNC client.
All the previous mobile software we did (and the 3D Pets series we did for the PC) was done using my own code - in the case of 3D Pets almost entirely in assembler - I even have my own blitter API that I wrote for (Microsoft and Symbian) ARM devices entirely in assembler (like others that were available but faster and now just as obsolete).
I haven't yet even looked at the source for Mandelbulber because I have no idea what SVN is (I'm not much good with anything not designed by myself or at least when I haven't written the equivalent myself).

Don't worry about "my head", as long I don't go back on the wacky backy I'll be fine provided I can start forcing myself to get more sleep !!

Make that just 3 values per bit of horizontal coordinates :)