Talonman

I found this post interesting...
http://forums.nvidia.com/index.php?showtopic=108761

'avidday'
I don't think so. Both NVIDIA and ATI are supplying special beta drivers with OpenCL support, but I don't believe either company have OpenCL support in their current release drivers. Further to that, I don't believe it is possible to build a single application that can run on either flavour of hardware. You can take OpenCL code and compile it with either vendor's SDK and it will probably work, but the back end code and support libraries are completely different and incompatible. Further to that, both flavours of OpenCL are really in beta and the performance and capabilities of both appears to be inferior to either vendors proprietary GPU compute environments (ie. CUDA or Stream).

It gave me the impression that 1 code may work on both GPU's, but 1 GPU vendor will loose real bad in performance.

It also confirmed for me that CUDA would be faster than OpenCL.

Sorry, didn't read through the whole thread but you are probably onto something here, more detail here from AnandTech in their OpenCL update published some time ago:
 
http://www.anandtech.com/weblog/showpost.aspx?i=648

AnandTech
OpenCL ICD: Coming Soon
 
 
Unfortunately matters are made particularly complex by the fact that on Windows and Linux, writing an OpenCL program right now requires linking against a vendor-specific OpenCL driver. The code itself is still cross-platform/cross-device, but in terms of compiling and linking OpenCL has not been fully abstracted. It’s not yet at the point where it’s possible to write and run a single Windows/Linux program that will work with any OpenCL device. It would be the equivalent of requiring an OpenGL game (e.g. Quake) to have a different binary for each GPU vendor’s drivers.
 
 
The solution to this problem is that OpenCL needs an Installable Client Driver (ICD), just like OpenGL does. With an ICD developers can link against that, and it will handle the duties of passing things off to vendor-specific drivers. However an ICD isn’t ready yet, and in fact we don’t know when it will be ready. NVIDIA - who chairs the OpenCL working group - tells us that the WG is “driving to get an ICD implementation released as quickly as possible”, but with no timetable attached to that. The effort right now appears to be on getting more OpenCL 1.0 implementations certified (NV is certified, AMD is in progress), with an ICD to follow.

 
As is the case more often than not, the promise of an "open standard panacea" falls short of expectations.  While OpenCL does provide a common language across platforms it still relies on vendor-specific support and an additional abstraction layer for full performance realization.   If the apps author used ATI's SDK and compiler (which he must've since he developed and debugged with a 4800 part) that most likely explains why performance is so abysmal on Nvidia parts and essentially runs in software on the CPU.  You will probably have to have the author or someone else with Nvidia hardware and sufficient programming knowledge re-write the app or if you know which headers need to be abstracted out, that might save some time, then recompile with Nvidia's OpenCL SDK.

Check out post #4 by pcchen, that basically reiterates the AnandTech blurb on ICDs.  OpenCL code still needs to link to vendor specific .dlls to tell the hardware how to run the code.  I'm guessing when you compile with any specific vendor's compiler it inserts the necessary links, but when run on another vendor's hardware, those links obviously don't work, so you need an ICD to abstract out those headers and replace them with the necessary Nvidia links/headers. 

Np Talonman, just picked up on your lead, hopefully it helps you find a solution.  Now you just need to bug Fred or someone else to re-write or re-compile the app for Nvidia hardware. 

Also in that post you linked, #6 from iAPX is also covered in that AT article about Apple's OpenCL implementation and why it actually works for both GPU vendors and CPU without any specific ICD:

AnandTech

Meanwhile Apple, in the traditional Apple manner, has simply done a runaround on the whole issue. When it comes to drivers they shipped Snow Leopard with their own OpenCL CPU driver, and they have GPU drivers for both AMD and NVIDIA cards. Their OpenCL framework doesn’t have an ICD per-say, but it has features that allow developers to query for devices and use any they like. It effectively accomplishes the same thing, but it’s only of use when writing programs against Apple’s framework. But to Apple’s credit, as of this moment they currently have the only complete OpenCL platform, offering CPU+GPU development and execution with a full degree of abstraction.

pcchen wrote:

To my understanding, there shouldn't be a difference (at least on Windows) between using AMD or NVIDIA's SDK. Basically the cl.h is almost the same (they are different in only one line, which is a comment). The opencl.lib files are different, but they both linked to opencl.dll, with completely the same functions and calling conventions. So there shouldn't be any benefit from recompiling with different SDK.

I think that is why we are seeing the same low performance for OpenCL on both the DirectCompute benchmark and SmallptGpu.   

Thanks for the link Nahte27... 

"This is about 45 time faster than SmallptCPU running on an Intel Q6600. Talonman has achieved some really high number of samples (on a very simple scene however) by using one of the 2 GPUs available with a NVIDIA GTX 295


114,000,000 of samples/sec is really an high number." 

I might get some street cred as a BETA Tester! 

I'm thrilled, thanks Luxrender....

Thanks,

I'm just glad we may have helped to speed up it's development. 

Do you see a link to download that version?

I can't seem to find it. 

It looks good.

I just asked Dave... 

http://forum.beyond3d.com...hp?t=55913&page=11

Nahte27

So, I've been following this thread for a while now. Interesting stuff IMO.

Today I was downloading some updates from the Luxrender website and I saw this: Luxrender and Open CL

I didn't realize that one of the reasons for this Open CL project is an Open CL powered version of Luxrender! Very cool stuff with a lot of potential!

Keep up the good work guys!

Dave has a message for you. 
http://forum.beyond3d.com...mp;t=55913&page=12

"Ah, thanks, could you write there to Nathe72 that the ".py" file is a ".ply" and there is an SmallLuxGPU exporter for Blender 2.5 available here: http://www.luxrender.net/forum/viewt...hp?f=34&t=3420

LuxrenderGPU uses instead the same .lxs file of Luxrender".

well, hook us up with some cool renders plz.