how to direct kernel calls to specific devices?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello:
I have two devices installed, on my Fedora 8, x86_64 machine. In theory, the 8600 GTS is used to run the monitor, and the GTX is for CUDA programming.
Device 0: “GeForce 8600 GTS”
Device 1: “GeForce GTX 280”

Can anyone tell me how I can direct CUDA calls to one or the other device? I tried using -arch ‘GeForce 8600 GTS’ with nvcc, but of course that did not work. How do you direct calls to specific devices?

Thanks, John

2

You have to call cudaSetDevice(int devicenum) to select a device before any other CUDA calls. Otherwise, the first CUDA function you call will pick device 0.

3

Thanks!!! Worked like a charm.