Figuring out which device is used by X11.

Accelerated Computing CUDA CUDA Programming and Performance
1

Is there a way to figure out which device is used by X11 using the CUDA API or something and then tell CUDA to use the other one (assuming there are two GPUs)

2

You need something like this:

int deviceCount;

cudaGetDeviceCount(&deviceCount);

for (int dev = 0; dev < deviceCount; ++dev)

{

	cudaDeviceProp deviceProp;

	cudaGetDeviceProperties(&deviceProp, dev);

	if (deviceProp.kernelExecTimeoutEnabled > 0)

	{

		// This device is running a display

	}

	else

	{

		// This device is not.

		// To use the first device that is not running a display:

		cudaSetDevice(dev);

		break;

	}

}

The device running the display will have kernelExecTimeoutEnabled set,

which stops a CUDA kernel freezing the display by running too long.

It is 5 seconds by default, I think. The other device will have this set to 0.

Then use cudaSetDevice() to choose which to use.

Of course if you are running two displays …

3

Didn’t know about the kernelExecTimeoutEnabled property. Thx for your quick replay :)