Multiple devices: Should you do cudaSetDevice() before every cuBLAS call?

Accelerated Computing CUDA CUDA Programming and Performance
1

http://docs.nvidia.com/cuda/cublas/#cublas-context says:

However, it doesn’t say whether the thread should make sure that the current device ID is correctly set before every cuBLAS library call. Should it?

I find that if it doesn’t, cuBLAS returns error code 13, and works fine otherwise.

2

If there was an earlier discussion about this, I can’t find it.

3

Never mind (bug in my own code)