Cuda to OpenCL! Read this! Tweet

[P!X3L]

Hello everyone,

I just came across this article, thought you might find this interesting.
http://www.vizworld.com/2010/03/switch-cuda-opencl-easily-swan/?utm_source=rss&utm_medium=rss&utm_campaign=switch-cuda-opencl-easily-swan

Translates CUDA kernel source-code to OpenCL

If this works with Octane....Man that would be sweet!!

Update: Also, check this out: http://www.anandtech.com/show/2977/nvidia-s-geforce-gtx-480-and-gtx-470-6-months-late-was-it-worth-the-wait-/6
Nvidea cards work extremely well for opencl, this means there are only advantages of switching over.
More customers

Swan website: http://www.multiscalelab.org/swan

Radiance, please check this out

Keen to know what everyone thinks, thanks.

[Maryus3D]

From what I know CUDA is more mature than OpenCL. It's true that OpenCL performs very well with nvidia cards but you'll have more performance from CUDA. The big advantage of OpenCL is that you can run it on multi platform. Anyway interesting article over there .

[P!X3L]

Thanks for your input Maryus3D

I have heard the same from various sources yes.
However, if this "Swan" program can make it really easy to make the switch; why not offer Octane in two flavors? Cuda and opencl? This way everybody wins!

[radiance]

Hi,

Sorry but it's still not the right time.
If we switch to opencl now, you'll see area lights appear sometime in 2011.
also our cuda code will run fastest, much faster than cuda code translated to opencl as opencl is slow. (even though it's a bit faster on nvidia/opencl).

i've ben aware of these new techs for a while now, incl. goose.

Radiance

[P!X3L]

Thanks for your reply Radiance,

Good to know that you have been researching these possibilities. I guess nothing is ever as simple as it seems
(especially if someone doesn't know anything about code).

I am referring to me of course hehe.

[Hawker]

it sounds like you want to learn to rad how make programs...

and, you know, stuff like this are not very good, it makes a bad code, nothing beats the total handmade.

[salvation]

from the link http://www.anandtech.com/show/2977/nvid ... he-wait-/6

NVIDIA has confirmed it - the GTX 400 series' FP64 performance is capped at 1/8th (12.5%) of its FP32 performance, as opposed to what the hardware natively can do of 1/2 (50%) FP32. This is a market segmentation choice - Tesla of course will not be handicapped in this manner. All of our compute benchmarks are FP32 based, so they remain unaffected by this cap.

that's unfair .. ok octane use fp32 but its unfair

[radiance]

from the link http://www.anandtech.com/show/2977/nvid ... he-wait-/6

NVIDIA has confirmed it - the GTX 400 series' FP64 performance is capped at 1/8th (12.5%) of its FP32 performance, as opposed to what the hardware natively can do of 1/2 (50%) FP32. This is a market segmentation choice - Tesla of course will not be handicapped in this manner. All of our compute benchmarks are FP32 based, so they remain unaffected by this cap.

that's unfair .. ok octane use fp32 but its unfair

the good thing is that we do support Geforce (even a first gen / old 8000 series) and endorse them, we won't turn on geforce and force people to buy quadro or tesla cards like other companies do.

Radiance

[havensole]

This is one of the great things with Octane versus many other rendering software. People of any walk of life can pick it up without hardware limitations or price limitations. A simple student hobbyist or a professional studio can venture into the world of gpu rendering without the worry of needing ridiculously expensive systems.