Page 1 of 1

Selecting a GPU to run Cula Sparse on

PostPosted: Thu Jul 30, 2015 5:30 pm
by reilya

I am having trouble selecting the device on which to run Cula Sparse on. I have two identical GPUs on my node. I have tried to set deviceId=1 in the PlatformOptions to tell the code to use the second GPU, but no matter what I do Cula Sparse seems to want to use the first one.

When I try to use the CudaPlatform in the plan, it runs correctly in the first GPU (even though I wanted to use the second one). When I use the CudaDevicePlatform the solver fails before it starts because the matrices have been correctly allocated in the second GPU, but the solver tries to run in the first one.

Is there anything else I need to set to select the GPU?

Any help is appreciated