I have seen that now it exists a gdb version for cuda, which is a very interesting feature.
I tried it, but it doesn’t work under Xorg. Like I have a gtx 295 with two gpu, I would like to know whether it exists some way to use one gpu for X window system and the other for executing kernel and cuda-gdb ?
Currently I use cudaSetDevice(1), but this way doesn’t work (or I don’t use it properly).
Just make sure X is not configured to run on one of the cards and use the nvidia-smi utility to set the display card to be compute prohibited. After that it should just work (I have a development box with a pair of GTX275s set up like this and it have never given me any trouble).