Archives Discussions

[yaun]

Hi all,

as far as I know an ATI GPU is accessed via the X server.

Given that, is it possible to run an application on my host A which uses the 

ATI GPU on another host B by just setting the DISPLAY environment variable

to host_B:0 ?

Thanks for your help.

Best

Sebastian

[omkaranathan]

Sebastian,

Please refer the following KB artilce.

http://developer.amd.com/support/KnowledgeBase/Lists/KnowledgeBase/DispForm.aspx?ID=19

[yaun]

Thanks for the response.

I read the document. However, it does not seem to describe the

scenario that I am looking at. Let me rephrase my question:

* Host A has an ATI GPU and X server running.

* Host B (e.g. my laptop) has *no* ATI GPU but is the location where my application is to run

* The application however uses the CAL for acceleration.

* What I'm trying to do is to run the CAL application on Host B (my laptop)

and since my laptop has no GPU for acceleration, the program should

offload the CAL computations transparently to Host A by simply

looking at the value of the DISPLAY variable which is set to Host B (which has the GPU)

Hopefully, this more clearly describes what I want to achieve.

I just would like to know if the DISPLAY variable can instruct the CAL library to look for the GPU

on another host.

Best

Sebastian

[nou]

i do not think it is possible.

[cjang]

Could you run your application remotely? So you could ssh from host B to host A and run there. My GPU compute server is almost always headless. I start the X server and run applications remotely over the network.

I agree with nou, what you are trying to do with the DISPLAY variable almost certainly won't work. But it sounds like what you want to do is very similar to my configuration. I typically work off laptops and access a compute server with discrete GPUs over the network.