CODEWITHSUNDEEP
macoscudansight
James Rome
James Rome

Reputation: 71

How to use two cuda cards and nsight Eclipse on a Mac

I finally have two nvidia cards in my Mac so that I can use Nsight Eclipse to debug CUDA code. But clearly I have not configured things properly because I only get a breakpoint into assembler, and then my Mac hangs if I continue. I think it is trying to run the CUDA program on the card used by the display, rather than the other card. I also get messages that it cannot find the symbols.

So how do I set things up properly? Thanks, Jim

Upvotes: 0

Views: 160

Answers (2)

Eugene
Eugene

Reputation: 9474

You can select the device to use for debugging in Nsight preferences. You can also set it for the debug configuration. Note that Nsight also honours CUDA_VISIBLE_DEVICES variable Robert mentioned.

Upvotes: 1

Robert Crovella
Robert Crovella

Reputation: 152143

perhaps the easiest solution might be to use the CUDA_VISIBLE_DEVICES environment variable which is described here Declare it in the session before you launch nsight EE, and that should be sufficient. You will have to experiment with it to find the right "mask" to select the GPU of interest.

The symbol messages are a different issue, I think. Are you trying to debug the debug project in nsight eclipse? That should create the symbols for you.

Upvotes: 1

Related Questions